# Erman Yaman > Erman Yaman is an AI-native product leader in Austin, TX who combines enterprise product leadership with hands-on AI engineering. As Senior Product Manager, eCommerce & UI/UX at McLane Company, a Berkshire Hathaway company, he led product and UI/UX for McLane360 from concept to launch in six months; it processed tens of millions of dollars in orders in its first six weeks. He founded ValuFact and built it solo, end to end: an AI financial-intelligence platform covering 8,000+ public companies, with deterministic SEC data pipelines, Google Gemini and open-weight LLMs in production, multi-stage filing synthesis and machine-learning models he trained himself, governed by pre-registered acceptance gates and immutable prediction ledgers. 15 years, 14 products taken to market. Key facts: - Name: Erman Yaman - Positioning: AI-native product leader who builds trained models, LLM systems with evals, agentic workflows and deterministic data pipelines - Current role: Senior Product Manager, eCommerce & UI/UX at McLane Company (a Berkshire Hathaway company), since February 2026 - Also: Founder, Product & Engineering Lead of ValuFact (https://valufact.com), an AI-native financial-intelligence platform he built solo, since October 2025 - Location: Austin, TX, United States - Focus areas: AI-native product development; Generative AI products; Large language model (LLM) integration; AI evaluation and governance; AI-assisted prototyping; LLM evaluation and model selection; Agentic AI workflows; Machine learning model training; Small language models and embeddings; AI cost optimization; Deterministic data pipelines; SEC EDGAR and XBRL data; US GAAP financial data standardization; Product strategy; Product management; B2B eCommerce; Enterprise SaaS; UX design; Product operations; Financial data products - At a glance: 15 yrs (Building products); 14 (Products taken to market); 1–2 wks (UI design cycles with AI prototyping); 20% (Of revenue from a practice I built) - Education: Master of Advanced Management, Yale School of Management (2018; Honors in seven courses, including Digital Strategy and Management of Software Development); MBA, Koç University Graduate School of Business (2017; Ranked 1st in class, 4.00 GPA; Class President and Student Representative); BA, Economics, Anadolu University (2008) - Contact: erman@yaman.us · LinkedIn https://www.linkedin.com/in/ermany/ - Résumé (JSON Resume): https://erman.yaman.us/resume.json - Availability: Open to conversations about senior AI product leadership roles. Not taking consulting work. What sets him apart: - **Leads products and ships them himself.** Built ValuFact solo, end to end (strategy, UX, architecture, data pipelines, the AI layer and production operations on FastAPI, PostgreSQL and a systemd worker fleet), and personally built the responsive functionality that shipped to production in McLane360. - **Trains models, not just prompts.** Trained a small language model with embeddings that maps patterns across SEC filings, an insider-trade intent model, score and signal prediction models and a GAAP concept classifier; earlier, a linear regression price model on ~27K car listings (2022). - **Rules first, models where they earn it.** A rule over SEC transaction codes labels 97% of insider filings at no cost, leaving a trained model one small job; 6-phase cleaning cuts filings 60–85% before the LLM stage; Gemini and open-weight models run in production under usage-based cost controls. - **Sets the bar before the result is in.** Pre-registered acceptance gates, masked holdouts with a logged one-time reveal, random null controls, point-in-time backtests and immutable prediction ledgers graded by code; the Pattern Map stays admin-only until it passes a pre-registered retrodiction test. - **Enterprise delivery at speed, with commercial results.** Took McLane360 from concept to launch in six months (tens of millions of dollars in orders in its first six weeks); AI-assisted code prototyping cut UI design cycles from months to one to two weeks; at Frumatic, built a UX service line that produced 20% of 2022 revenue. AI stack: - In production inside ValuFact: Google Gemini, Open-weight models - Used to build the platform, not inside the product (cost): OpenAI GPT, Anthropic Claude ## AI-native capability evidence - **Machine-learning models he trained** ([ValuFact case study](https://erman.yaman.us/projects/valufact/), [Price-prediction case study](https://erman.yaman.us/projects/price-prediction-market-insights/)): Pattern Map: defined and trained a small language model with embeddings that detects and classifies emerging and declining patterns in SEC filings and maps how they relate to companies and to each other; it stays admin-only until it passes a pre-registered retrodiction test on the live pipeline. Insider-trade intent model for the filings a deterministic rule over SEC transaction codes cannot settle (the rule labels 97% of insider filings at no cost). Score and signal prediction models, graded by code against an immutable prediction ledger; nothing is re-scored after the fact. GAAP concept classifier working alongside the rule-based and FASB layers of a 4-layer engine that maps 17,738 GAAP concepts to standardized line items. Linear regression price model on ~27K car listings (Frumatic price-prediction toolkit, 2022). - **LLM systems in production** ([ValuFact case study](https://erman.yaman.us/projects/valufact/), [valufact.com](https://valufact.com)): In production inside ValuFact: Google Gemini and open-weight models. OpenAI GPT and Anthropic Claude were used while building the platform (AI-assisted and agentic engineering), not inside the product, for cost reasons. Filing-type-specific prompt architectures with rigid JSON output schemas across six SEC filing types (10-K, 10-Q, 8-K, Form 4, 20-F, 6-K). Token counting to choose the processing strategy, multi-chunk synthesis for large filings and strict missing-data handling so the model never invents figures. Model choice by measured quality and cost; usage-based AI cost controls are built into four-tier subscription packaging. - **Agentic and multi-step AI workflows** ([ValuFact case study](https://erman.yaman.us/projects/valufact/), [McLane360 case study](https://erman.yaman.us/projects/mclane360/)): Multi-stage filing synthesis: chunk-level analyses of large filings are combined into one schema-bound result. Autonomous monitoring agents and research/analysis agents run ValuFact's internal operations (behind the scenes, not a user-facing feature). Agentic engineering: builds and ships with AI coding agents (OpenAI GPT, Anthropic Claude). At McLane, introduced an AI-assisted, code-based prototyping practice that cut UI design cycles from months to one to two weeks. - **Evaluation and model governance** ([ValuFact case study](https://erman.yaman.us/projects/valufact/)): Pre-registered acceptance gates: every analytical feature has a pass mark written down before it is measured. Masked holdouts with a one-time, logged reveal, and random null controls alongside every sweep. Point-in-time backtesting that replays scores exactly as they stood on each date, under an anti-overfitting experiment protocol. Immutable prediction ledgers: predictions are written once with a start point and deadline and graded by code. Full configuration audit trail (who, why, before and after) with restore to any version. - **Deterministic, automated data pipelines** ([ValuFact case study](https://erman.yaman.us/projects/valufact/)): Six-stage SEC filing pipeline: RSS detection every 10 minutes with trading-calendar awareness, rate-limited EDGAR download, 6-phase cleaning that cuts file size 60–85%, token counting, AI analysis and pre-computed storage. Rules first, models where they earn it: wherever a deterministic rule measures better than a model, the rule ships, with deterministic fallbacks wherever AI does not outperform rules. 8,000+ public companies covered; standardized financials for 7,000+ from 17,738 GAAP concept mappings, pulled directly from SEC EDGAR with no third-party aggregator. 18-ratio engine and six composite scores (Quality, Valuation, Solvency, Growth, Efficiency, Overall) recomputed nightly. - **Production engineering** ([ValuFact case study](https://erman.yaman.us/projects/valufact/), [McLane360 case study](https://erman.yaman.us/projects/mclane360/)): Built and runs the production stack solo: Python/FastAPI, PostgreSQL, a systemd worker fleet and encrypted off-box backups. Personally built the responsive functionality that shipped to production in McLane360, extending it from mobile-only to tablet and desktop. - **AI-native product leadership in the enterprise** ([McLane360 case study](https://erman.yaman.us/projects/mclane360/), [Business Wire](https://www.businesswire.com/news/home/20260901909156/en/), [McLane announcement](https://www.mclaneco.com/mclane-company-debuts-mclane360-showcases-new-products-and-expanded-foodservice-innovation-at-mclane-engage/)): Led product and UI/UX for McLane360 at McLane Company (a Berkshire Hathaway company) with one Product Owner and two UI/UX leads, from concept to launch in six months; tens of millions of dollars in orders in its first six weeks. Runs biweekly steering and quarterly roadmap planning with senior business, technology, sales and marketing leaders. Established an eight-phase UX delivery framework and design-system standards. At Frumatic, directed 11 client products across four countries and built a UX service line that produced 20% of 2022 revenue (54% close-to-quote). ## Experience - **Senior Product Manager, eCommerce & UI/UX, McLane Company** (Feb 2026 – Present, Austin, TX): Own product strategy, roadmap and customer experience for McLane360, the unified B2B customer platform, leading a team of one Product Owner and two UI/UX leads. - Turned a request to extend browser ordering into McLane360, one platform for promotions, ordering, delivery visibility and business insights, and took it from concept through beta to launch in six months. The platform processed tens of millions of dollars in orders within its first six weeks. - Run product steering and quarterly roadmap planning with senior business, technology, sales and marketing leaders, resolving trade-offs across workstreams. - Launched at McLane Engage 2026 with customer and teammate satisfaction both above 4 out of 5 in post-event surveys, beating the target set before the show. - Established an eight-phase UX delivery framework, design-system standards and an AI-assisted, code-based prototyping practice that cut UI design cycles from months to one to two weeks. - Designed a vendor-agnostic product-data architecture that forms the business case for future eCommerce investment. - **Founder, Product & Engineering Lead, ValuFact** (Oct 2025 – Present, Austin, TX): Conceived and built an AI-native financial-intelligence platform end to end: product strategy, UX, architecture, SEC data pipelines, APIs and production operations. - Covers 8,000+ public companies, with standardized financials for 7,000+ built on 17,700+ GAAP concept mappings. - Shipped AI filing analysis, 18 financial ratios, six composite scores, screeners, alerts, market intelligence and point-in-time backtesting; designed four-tier subscription packaging with usage-based cost controls. - Built model governance into the product: masked holdout testing, pre-registered acceptance gates, immutable prediction ledgers and a full configuration audit trail, with deterministic fallbacks wherever AI did not outperform rules. - Trained the platform's machine-learning models: a small language model with embeddings for the Pattern Map (admin-only until it passes a pre-registered retrodiction test), an insider-trade intent model, score and signal prediction models graded by an immutable prediction ledger, and a GAAP concept classifier. - Runs Google Gemini and open-weight models in production; used OpenAI GPT and Anthropic Claude to build the platform (AI-assisted and agentic engineering), not inside the product, for cost reasons. - Autonomous monitoring and research/analysis agents run the platform's internal operations; multi-stage synthesis turns large filings into one schema-bound analysis. - **Global Product Director, Frumatic** (2019 – Feb 2026, Austin, TX · remote): Directed product strategy and delivery for 11 client products across four countries, leading two Product Managers and two UI/UX Designers alongside distributed engineering teams. - Built and commercialized a UX service offering from the ground up (team, operating model and client proposition) that generated 20% of company revenue in 2022, with a 54% close-to-quote ratio on proposals. - Led products across fintech (open-banking and Gen Z banking apps), automotive dealership SaaS, healthcare and hospitality, growing annual project throughput 26%. - Standardized discovery, requirements and user-story practices across teams, raising sprint velocity 57%. - Led market analysis, customer research and the roadmap for a Nordic telecom conglomerate’s digital presence, driving a 103% increase in pageviews in seven months. - **Head of Product, Storia Inc.** (2019 – 2020, Istanbul, Turkey): Led the pivot to an interactive user-generated-content platform on a startup budget, refocusing the company on two markets. - Grew users 34% to 1.7M and pageviews 148% to 17.2M through cross-functional design and engineering delivery. - Designed a points-and-ranking system that lifted pageviews per session 313% and session duration 156% while cutting bounce rate 73%. - Narrowed operations to Turkey and Brazil, rebuilt the team around a lean model, and won investor buy-in for the new product strategy. - **Marketing Programs Manager, Tuyap Exhibitions Group** (2015 – 2017, Istanbul, Turkey): Identified, developed and commercialized B2B digital services, coordinating software, BI, IT and commercial teams. - Generated $1.5M in digital revenue in 2016 without capital investment. - B2B recommendation products improved Net Promoter Score by 29 points and customer satisfaction by 18% across two pilot exhibitions. - **Product Development Specialist, Kale Industries Holding Co.** (2014 – 2015, Istanbul, Turkey): Translated customer research across 12 cities into pricing and go-to-market strategy for two category-defining launches with ~$250M in projected revenue. - Represented the customer across the development and improvement of 40 products. - **Consumer Electronics Product Manager, Arena Computers Inc.** (2011 – 2014, Istanbul, Turkey): Owned market analysis, portfolio and pricing strategy for the consumer-electronics category. - Grew category revenue 41%, from $10.8M to $15.2M in two years, more than twice the company growth rate. Full timeline: https://erman.yaman.us/experience/ ## Case studies - [McLane360](https://erman.yaman.us/projects/mclane360/) (Launched August 2026 · Six-month delivery): Led eCommerce product and UI/UX for McLane360 with a team of one Product Owner and two UI/UX leads, shaping the ordering experience from the ground up and helping bring McLane’s unified customer platform to market. - [ValuFact](https://erman.yaman.us/projects/valufact/) (Oct '25 – Present): AI-native financial-intelligence platform that turns SEC filings into plain-English summaries, standardized financials, scores and screens for 8,000+ public companies. - [Frumatic DMS](https://erman.yaman.us/projects/frumatic-dms/) (Jan '22 – Feb '26): Smart, ultra-flexible, and connected SaaS platform that streamlines auto dealership management. - [A Banking App for GenZ*](https://erman.yaman.us/projects/banking-app-genz/) (Apr '22 – Aug '22): A banking app for GenZ that encourages positive money habits and rewards those who embrace them. - [Price Prediction & Market Insights](https://erman.yaman.us/projects/price-prediction-market-insights/) (Feb '22 – Mar '22): Analytics toolkit that accurately estimates car market values and displays competitor inventory details. - [Mental Wellbeing Platform*](https://erman.yaman.us/projects/mental-wellbeing-platform/) (Nov '20 – Feb '22): Mental well-being apps with conversational assessments, knowledge base, and live chat features. - [Donations App](https://erman.yaman.us/projects/donations-app/) (Jan '20 – Oct '21): Open banking app that rounds up spending and automatically donates to preferred charities. - [3D Cabinet Designer](https://erman.yaman.us/projects/3d-cabinet-designer/) (Aug '20 – Dec '21): Browser-based 3D cabinet configurator detecting part compatibility and sending orders to the factory. - [HORECA Booking Platform](https://erman.yaman.us/projects/horeca-booking-platform/) (Apr '22 – Dec '22): Suite of web and mobile apps for restaurants, cafes and clubs to manage bookings and customers to make reservations. - [Camping Platform](https://erman.yaman.us/projects/camping-platform/) (Feb '22 – Mar '22): A camping platform that allows camp-sites to manage their inventory and campers to book spaces effortlessly. - [Beauty Services Platform](https://erman.yaman.us/projects/beauty-services-platform/) (Feb '22 – Mar '22): Web platform allowing beauty service providers to manage bookings, services, availability, and payments. ## Corroboration - [McLane Company debuts McLane360, showcases new products and expanded foodservice innovation at McLane Engage](https://www.mclaneco.com/mclane-company-debuts-mclane360-showcases-new-products-and-expanded-foodservice-innovation-at-mclane-engage/) (McLane Company; about McLane360) - [McLane360 debuts at McLane Engage](https://www.businesswire.com/news/home/20260901909156/en/) (Business Wire; about McLane360) - [McLane Co. launches new digital platform for customers](https://cspdailynews.com/technologyservices/mclane-co-launches-new-digital-platform-customers) (CSP Daily News; about McLane360) - [McLane360 on the App Store](https://apps.apple.com/us/app/mclane360/id6754037170) (Apple App Store; about McLane360) - [McLane360 on Google Play](https://play.google.com/store/apps/details?id=com.mclaneco.apps.mobile.grocery.smarthandheld.mclane360) (Google Play; about McLane360) - Tanya Coutray, Business Unit CIDO, McLane Company: "...with special recognition to Lance, Erman, and Sri Putumbaka for their leadership and their team members contributions throughout the journey." (Reflection on the McLane360 launch at McLane Engage 2026; [source](https://www.linkedin.com/posts/tanyacoutray_mclaneengage-mclane360-digitalleadership-activity-7505075756230717440-8Yih)) ## Achievements - [6 months Concept to Launch](https://erman.yaman.us/projects/mclane360/): McLane360: From Product Vision to Customer Launch in Six Months. Role: Senior Product Manager - [1,270% User Growth](https://erman.yaman.us/achievements/pivot-bootstrapping-storia/): A Pivot & Bootstrapping Story: 90K to 1.2M Visitors in One Year. Role: Head of Product - [73% Throughput Growth](https://erman.yaman.us/achievements/product-ops-frumatic-velocity/): Product Operations Management: 73% Velocity Increase in 6 Months. Role: Global Product Director - [103% Pageview Growth](https://erman.yaman.us/achievements/corporate-website-renewal/): Aligning an EVP and Seven Department Heads on One Roadmap: 103% Pageview Growth. Role: Strategy lead, team of four ## How I lead - [Product Strategy](https://erman.yaman.us/expertise/product-strategy/): I take products from concept to business case to market, decide where AI genuinely changes the product, and make the trade-offs that keep a roadmap honest. - [Product Operations](https://erman.yaman.us/expertise/product-operations/): I build the operating rhythm a product organization runs on: steering, planning, shared standards, visible delivery and AI-assisted ways of working. - [UX & Product Experience](https://erman.yaman.us/expertise/ux-product-experience/): I treat experience as the product: shared design standards, AI-assisted prototypes built in code, and satisfaction measured against a target. ## Skills - AI-native product development: AI-assisted prototyping, AI-assisted development, LLM product integration, Structured outputs, Evaluation & acceptance gates, Model governance, AI cost controls, Deterministic fallbacks - Product & commercialization: Product strategy, Business cases, Pricing & packaging, Multi-year roadmaps, Go-to-market, Customer & market research - Operating rhythm & teams: Team building, Steering committees, Quarterly planning, Delivery frameworks, Design-system standards, Executive alignment - Tools & stack: Python / FastAPI, Next.js / TypeScript, PostgreSQL, SQL, REST APIs, HTML / CSS / JavaScript, Jira, Confluence, Figma, Miro, Google Analytics, Firebase, Google AI Studio, Cloudflare, VS Code, WordPress - AI & ML engineering: Machine learning model training, Small language models and embeddings, LLM evaluation and model selection, Agentic AI workflows, Multi-stage LLM pipelines, Structured outputs (JSON schemas), Pre-registered evaluation and acceptance gates, Holdout testing and point-in-time backtesting, Model governance and audit trails, Deterministic data pipelines, AI cost optimization, SEC EDGAR and XBRL data, US GAAP financial data standardization, Python / FastAPI, PostgreSQL ## Insights - [Start With the Rule, Not the Model](https://erman.yaman.us/insights/start-with-the-rule-not-the-model/) (October 28, 2026): When I built ValuFact, the best AI decision I made was keeping a model away from 97% of a job. A plain rule did it for free. Here is how I decide where a model actually earns its place... - [When Building Gets Cheap, Deciding Gets Expensive](https://erman.yaman.us/insights/when-building-gets-cheap-deciding-gets-expensive/) (October 28, 2026): AI took my team from months of design cycles to a week or two. I expected everything to get faster. It did not. The hard part did not shrink, it moved. Here is where it went... - [Write the Pass Mark Down Before You Look](https://erman.yaman.us/insights/write-the-pass-mark-before-you-look/) (October 28, 2026): Imagine a student who grades their own exam and decides the pass mark after seeing the score. They would never fail. That is how a lot of teams evaluate launches, experiments and AI models... - [Strategy: The Art of Not-Choosing](https://erman.yaman.us/insights/strategy-the-art-of-not-choosing/) (April 2, 2023): Imagine yourself on a small boat with ten holes leaking water, but you only have enough material to cover seven. The key to reaching shore lies in deciding which holes to cover... - [An Introductory Guide to UX Workshop Activities](https://erman.yaman.us/insights/ux-workshop-activities/) (March 14, 2023): User Experience workshops bring together designers, stakeholders, and users to collaborate on creating and refining exceptional digital products... - [Key Characteristics of a Top-Notch Software Development Company](https://erman.yaman.us/insights/key-characteristics-software-dev-company/) (March 1, 2023): Drawing on my experience collaborating with software development companies for over a decade, here are my insights on what to consider when seeking a services provider... ## Case study details ### ValuFact Source: https://erman.yaman.us/projects/valufact/ ValuFact turns SEC filings — the dense, 200-page documents public companies are required to file — into plain-English summaries, standardized financial data, scores and screens. Automatically, within minutes of publication, for more than 8,000 companies. It is live at [valufact.com](https://valufact.com). ValuFact is an independent venture I founded and built end to end on my own time: strategy, UX, architecture, data pipelines, the AI layer and production operations. Building it alone was deliberate. It is how I learned first-hand what AI-native product development demands — data quality, evaluation, cost control and governance — so I can set credible standards for the teams I lead. #### What did I do in this project? - Identified the gap between institutional-grade financial data tools ($20K+ a year) and what individual investors can access - Conceived, designed and built the product end to end, from concept and packaging to architecture and operations - Engineered real-time SEC monitoring that detects new filings within minutes of publication - Built a 6-phase document cleaning pipeline that turns raw SEC filings (HTML, XBRL, SGML) into clean, structured text — cutting file size 60–85% while preserving the meaningful content - Designed filing-type-specific AI prompt architectures with rigid JSON output schemas for consistent analysis across 10-K, 10-Q, 8-K, Form 4, 20-F and 6-K filings - Developed a 4-layer financial concept classification engine mapping 17,738 GAAP concepts to standardized statement line items - Built an 18-ratio analysis engine and six composite scores with data-quality tracking, trend history and plain-English explanations - Designed four-tier subscription packaging with usage-based AI cost controls - Built model governance into the product: pre-registered acceptance gates, masked holdouts, immutable prediction ledgers and a configuration audit trail - Run the production stack — FastAPI, PostgreSQL and a systemd worker fleet — with encrypted off-box backups #### The Problem Every quarter, thousands of publicly traded companies file detailed financial reports with the SEC. These filings are the single most authoritative source of truth about a company's financial health — but almost no one can actually use them. - **They're enormous.** A typical 10-K filing runs 200+ pages of dense legal and financial text. - **They're written in accounting jargon.** The same metric might appear under 17,000+ different naming conventions across companies. - **They're inconsistent.** Apple's balance sheet doesn't look like Goldman Sachs's — even though they report to the same regulator. - **There's no standard summary.** Making sense of raw filings requires financial expertise, context, and hours of time. The tools that solve this — Bloomberg, Refinitiv, FactSet — cost $20,000–$25,000 per year. ValuFact closes this gap. #### AI-Powered Filing Analysis When a company files a document with the SEC, ValuFact's backend detects it, downloads it, and runs it through a proprietary multi-stage processing pipeline. What comes out is a structured, human-readable analysis. ##### What Users See Every filing is transformed into a set of structured, readable views: - **Headline summary** — One plain-language sentence on what the filing says - **What happened** — The facts and figures, point by point - **Why it matters** — What the results mean for the company - **Drivers and risks** — What drove the results, and the risks the filing itself names - **What comes next** — What to watch from here - **Section by section** — Every part of the filing summarized on its own (22 sections for Apple's 10-K) - **The numbers in this filing** — Headline figures from the filing's own XBRL against the year before, with links back to the original on sec.gov ##### Before & After: Raw SEC Filing → ValuFact Summary The raw SEC filing (left) vs. what ValuFact produces (right) — the same document, transformed into structured, actionable intelligence. Raw SEC Filing ValuFact Analysis ##### Before & After: Raw Financial Data → Standardized Statements Raw financial data buried in SEC filings (left) vs. clean, standardized, and comparable statements ValuFact produces (right). Raw SEC Data ValuFact Standardized #### The Multi-Stage Processing Pipeline Every filing goes through a purpose-built 6-stage pipeline: 1. **Detection** — Continuous SEC RSS monitoring every 10 minutes with trading calendar awareness 2. **Download** — Rate-limited retrieval from EDGAR, paced to respect SEC access limits 3. **Cleaning** — A 6-phase document transformation: XBRL truncation, header extraction & section marking, table extraction to CSV, HTML cleaning, section detection, and normalization 4. **Token Counting** — Precise token measurement via API to determine optimal processing strategy 5. **AI Analysis** — Filing-type-specific prompt architectures with structured JSON schemas, multi-chunk synthesis for large filings, and strict handling of missing data to eliminate hallucinations 6. **Storage & Delivery** — Analysis is pre-computed and stored so users get instant results with zero loading delay #### Financial Data Standardization ValuFact's 4-layer concept classification engine maps every one of 17,738 GAAP concepts to a standardized financial statement structure: 1. **Manual Overrides (100% confidence)** — Hand-crafted mappings for edge cases 2. **FASB Presentation Linkbase (100% confidence)** — Official accounting standards board guidelines 3. **FASB Schema Matches (95% confidence)** — Validated elements from FASB schema definitions 4. **Heuristic Rules (70% confidence)** — Pattern-based classification for remaining concepts The result: every company's financials are presented in an identical, comparable format — Balance Sheet, Income Statement, Cash Flow, and Statement of Stockholders' Equity. #### Financial Ratios & Metrics ValuFact automatically calculates 18 key financial ratios across five categories: - **Profitability** — ROA, ROE, Net Profit Margin, Gross Profit Margin, Operating Margin - **Liquidity** — Current Ratio, Quick Ratio, Cash Ratio - **Leverage** — Debt-to-Equity, Debt-to-Assets, Equity Ratio, Interest Coverage - **Cash Flow** — Operating Cash Flow Ratio, Free Cash Flow, Free Cash Flow Ratio - **Efficiency** — Asset Turnover, Inventory Turnover, Receivables Turnover Each ratio includes historical trend visualization, period-over-period comparisons, and plain-English explanations of what the numbers mean. #### Company Profiles & Explore Users can browse and search across 8,000+ companies, filter by sector or industry, and access dedicated company hubs showing recent filings with AI summaries, standardized financial statements, ratio dashboards, and key metrics with trend charts. #### What Makes ValuFact Different - **Not a ChatGPT wrapper** — Purpose-built infrastructure with a 6-phase cleaning pipeline, filing-type-specific prompts, rigid output schemas, and multi-stage chunk synthesis - **Direct from the source** — Every piece of data pulled directly from the SEC's EDGAR system — no third-party aggregator - **Transparency by design** — Every summary links to the original filing; every financial statement shows a coverage score - **Built for both audiences** — ELI5 summaries for regular investors, standardized multi-period financials for professionals #### Model governance and controls AI is core to ValuFact, so it is governed like any other production system. Nothing reaches users because a model sounds confident; it ships when it clears a bar that was set before the test. - **Pre-registered acceptance gates** — Every new analytical feature has a pass mark written down before it is measured. The cross-company Pattern Map, for example, stays admin-only until it passes a pre-registered retrodiction test on the live pipeline. - **Masked holdouts and point-in-time testing** — Backtests replay scores exactly as they stood on each date. Experiment windows are fixed at creation, the holdout is masked with a one-time, logged reveal, and random null controls run alongside every sweep. - **Immutable ledgers** — Predictions are written once, with their starting point and deadline, and graded by code. Nothing is re-scored after the fact. - **A full configuration audit trail** — Every change to scoring and theme settings is recorded with who, why, before and after, and any version can be restored. - **Rules first, models where they earn it** — Where a deterministic rule measures better than a model, the rule ships. For insider-trade intent, a rule over the SEC's own transaction codes labelled 97% of filings at no cost, leaving a model one small, well-defined job. #### Since launch The first release summarized filings and standardized financials. Since then ValuFact has grown into a research platform: - **Six composite scores** per company — Quality, Valuation, Solvency, Growth, Efficiency and Overall — recomputed nightly - **Screener, watchlists, holdings and threshold alerts** built on the 18 ratios - **Market & Economy page** with index proxies, the Treasury curve, economic indicators and earnings calendars - **Insider and congressional trade intelligence**, with each trade's outcome measured against the market - **Point-in-time backtesting** with an anti-overfitting experiment protocol - **Four-tier subscription packaging** — free, basic, professional and enterprise — with usage-based AI cost controls - **Encrypted off-box backups** and an indexed archive of every stored filing #### By the Numbers - **8,000+ companies** covered, **7,000+** with standardized financial data - **17,738 concept mappings** in the classification engine - **18 financial ratios** calculated automatically with trend analysis - **6 composite scores** per company, recomputed nightly - **6 filing types** supported with type-specific AI analysis (10-K, 10-Q, 8-K, Form 4, 20-F, 6-K) - **4 subscription tiers** designed around usage-based AI costs --- #### Appendix: Full 10-K Analysis Example ValuFact's complete AI-generated analysis of Apple's fiscal 2025 10-K, as users see it: the headline summary, what happened, why it matters, drivers and risks, what comes next, a section-by-section breakdown and the filing's own numbers. ### McLane360 Source: https://erman.yaman.us/projects/mclane360/ 6 months Vision to public launch 4+ / 5 Customer and teammate satisfaction $XXM In orders within six weeks Satisfaction results come from post-event surveys at McLane Engage 2026. #### A bigger opportunity than an ordering app The starting point was a request to extend browser-based ordering. I saw an opportunity to solve a broader customer problem: fragmented tools for discovering promotions, placing orders, and staying informed. That thinking helped shape McLane360, a unified customer platform bringing promotion ordering, delivery tracking, and business insights into one experience. As **Senior Product Manager, eCommerce & UI/UX**, I led the product direction and experience design for ordering, working with partners across the business and technology teams to deliver the public launch in six months. I led a team of one Product Owner and two UI/UX leads and owned the roadmap: product steering with senior business, technology, sales and marketing leaders, and quarterly roadmap planning. The milestone was McLane Engage 2026, where customers could use the product on the show floor. McLane featured the launch in its [official company announcement](https://www.mclaneco.com/mclane-company-debuts-mclane360-showcases-new-products-and-expanded-foodservice-innovation-at-mclane-engage/). #### My focus: make complex ordering feel simple For a retail buyer, placing an order can mean comparing promotions, understanding product details, and allocating purchases across multiple stores. I shaped the journey around three connected tasks: **discover, evaluate, and order**. - **Find the right deal.** Prioritized intuitive browsing, search, filtering, and Favorites so buyers could navigate the catalog with confidence. - **Make an informed choice.** Brought product information, imagery, pricing, and shipping details into the purchasing flow, kept current through catalog synchronization. - **Check out across stores.** Shaped a flexible cart and order-allocation experience, with checkout hardening and responsive design central to launch readiness. McLane360 product marketing images. Select a screen to view it at full size. #### Building the practice, not just the product A fixed date and a small team meant the way we designed had to change too. I established an eight-phase UX delivery framework and design-system standards, so every feature moved from discovery to developer handoff the same way. The biggest shift was prototyping in code. With AI-assisted development, the team built interactive HTML, CSS and JavaScript prototypes instead of static mockups, which cut UI design cycles from months to one to two weeks across the entire application. I stayed close to the build as well. I personally developed the responsive functionality that shipped to production, extending McLane360 from its original mobile-only scope to tablet and desktop. #### Six months. A fixed launch. Real customers. An event launch creates a clear test: the product has to work when customers arrive. I drove sprint priorities across checkout, Favorites, responsive design, catalog synchronization and launch readiness, keeping the ordering experience at the center of delivery decisions. Before Engage, we ran a structured beta with sales and marketing teammates to exercise the live ordering flow and reduce launch risk. Biweekly steering committees kept leadership and cross-functional partners aligned on scope and launch readiness as the team moved toward the show. On the floor, the work extended beyond software. Sales, merchandising, marketing, engineering, and support teams helped suppliers prepare promotions and customers get set up. Seeing buyers discover deals and complete orders was the most rewarding validation of the experience we had built. #### Customer validation beyond launch day The orders followed quickly: within its first six weeks, McLane360 processed **tens of millions of dollars in orders**. Post-event surveys from **both customers and teammates rated the launch above 4 out of 5**, beating the satisfaction target the team set before the show, with strong ratings for ease of navigating the app and finding products and information. The launch established a foundation for a broader digital relationship with McLane customers. Alongside the launch, I designed a vendor-agnostic product-data architecture so future eCommerce initiatives don’t depend on any single product-data vendor. It forms the business case for the next round of eCommerce investment. #### Recognition for a shared achievement In her reflection on Engage, Tanya Coutray, McLane’s Business Unit CIDO, recognized the Digital Customer Experiences team and called out our leadership: “...with special recognition to Lance, Erman, and Sri Putumbaka for their leadership and their team members contributions throughout the journey.” — [Tanya Coutray, on LinkedIn](https://www.linkedin.com/posts/tanyacoutray_mclaneengage-mclane360-digitalleadership-activity-7505075756230717440-8Yih) Chief Customer Experience Officer Vito Maurici also thanked me alongside the IT&D leaders and teams in his launch reflection. That recognition reflects a collective effort: organizational leadership across Digital Customer Experiences, the engineering team’s delivery, the teams behind delivery tracking and shared features, and the many business partners who made the launch possible. #### Explore the launch - [McLane’s official launch announcement](https://www.mclaneco.com/mclane-company-debuts-mclane360-showcases-new-products-and-expanded-foodservice-innovation-at-mclane-engage/) - [Business Wire: McLane360 debuts at Engage](https://www.businesswire.com/news/home/20260901909156/en/) - [CSP: McLane launches a new digital platform for customers](https://cspdailynews.com/technologyservices/mclane-co-launches-new-digital-platform-customers) - [McLane360 on the App Store](https://apps.apple.com/us/app/mclane360/id6754037170) - [McLane360 on Google Play](https://play.google.com/store/apps/details?id=com.mclaneco.apps.mobile.grocery.smarthandheld.mclane360) ### Frumatic DMS Source: https://erman.yaman.us/projects/frumatic-dms/ As Global Product Director at Frumatic, I built and led the product and UX team behind the DMS (two Product Managers and two UI/UX Designers, alongside distributed engineering) and owned the multi-year roadmap for the Vehicle Trading Platform ecosystem it anchors. DMS is an all-in-one digital solution designed to help auto dealers run their businesses with ease and efficiency. With the ability to automate processes, monitor employee performance, and leverage data to improve profitability, Frumatic's DMS can help auto dealers reduce operating expenses and improve the bottom line. #### What did I do in this project? - Researched the market for trends, opportunities, and risks - Collaborated with auto retailers to understand their workflow, needs, and pain points - Created the product vision, strategy, and roadmap for an ecosystem of digital products that enable auto retailers to run their businesses digitally, with minimum effort and operating expenses - Wrote product and task requirements, designed user-flows, drew wireframes, and outlined business logic to effectively communicate expected functionality across all stakeholders - Designed an ultra-flexible and connected information architecture that would allow the system to evolve quickly without losing reliability or stability - Created a scalable, reusable component system and utilized a front-end framework to enable fast development, maintenance, and improvement - Managed Product, UX, and Software Development teams to successfully deliver Version 1 #### Turning a capability into a revenue line The design discipline built for products like the DMS became a business in its own right. I turned Frumatic's in-house UX capability into a client-facing service line, covering the team, the operating model and the client proposition. In 2022 it generated 20% of company revenue, and proposals I contributed to closed at a 54% close-to-quote ratio. #### VTP & DMS DMS is the brains and nervous system of the Frumatic Vehicle Trading Platform ecosystem which also consists of: Analytics & AI, CRM, After Sales, and Listing CMS platforms. ##### Makes Life Easier Automate processes, monitor employee performance and leverage data to improve the bottom-line. ##### Connected Send and receive data from 3rd party systems with ease with our 3rd Party data-matching infrastructure. ##### Ultra-Flexible Our system is built to change and evolve to better serve your business needs and processes. #### Equalizer for Independent Auto Retailers Automotive is the largest yet least digitalized and the most fragmented retail market in the world. By providing cost-efficient solutions that enable our clients to operate their businesses digitally and with low operating expenses, we aim to level the playing field and help our clients succeed in a competitive marketplace. ##### Connect Effortlessly exchange data with third-party systems — vehicle information, market listings, financing applications, and insurance quotes. ##### Automate Automate your workflows to increase your company's efficiency and reliability while reducing operating expenses. Includes e-sign documents, automated accounting, expense tracking, and inspection & service task generation. ##### Monitor & Manage Built-in systems that make operations management easier than ever: to-do lists, communication tools, notification system, and document management. #### Make Smart Decisions - **Price Prediction** — Get accurate estimated market price when selling or buying cars - **Cost Calculations** — Depreciation, op-ex, and financial costs per vehicle - **Market Insights** — Automatically track competitor inventory and market trends - **Financial Performance** — Monitor sales, expenses, and profits per branch or sales rep ### Price Prediction & Market Insights Source: https://erman.yaman.us/projects/price-prediction-market-insights/ In this project, I developed a set of analytical tools that help auto dealers accurately estimate the prices of cars they buy or sell, get insights about their competitors' inventory, calculate depreciation, and find cars listed below market value. #### What did I do in this project? - Established an ethical and conscious way to scrape publicly available listing data (no copyrighted or personal information was scraped, throttled crawler requests) in the Finnish Automotive Market - Created a data structure that allows classifying and grouping listings per Make, Model, Trim, Model Year, Auto Retailer, Branch, and Sub-Branch - Built a linear regression model to define the relationship for each model generation between mileage, model year, and package (or certain keywords in the description) - Built a dashboard to visualize the results and allow users to search for specific information with ease #### Key Capabilities - **Inventory Intelligence** — Volume (quantity and €), make/model/trim/year distribution, average age, price, and mileage per branch - **Price Prediction** — Accurate estimated market price based on regression modeling - **Depreciation Calculator** — Per make, model, and trim - **Best Deals Finder** — Cars listed below calculated market value #### Demo The system was built on Google Looker Studio (formerly Google Data Studio), retrieving around 27K individual listings. The data is a snapshot from February 2022 and is not current. Reports are originally designed for web and are not mobile optimized.