# dataacrobat.com — Jay Ozer

> Portfolio of Jay Ozer, an AI engineering leader who builds production AI for document intelligence, autonomous agents and regulated workflows. Senior Manager, Applied AI at Doma Technology.

## Key facts

- Professional profile: https://www.dataacrobat.com/about
- Name: Jay Ozer
- Role: Senior Manager, Applied AI at Doma Technology (San Francisco, CA)
- Focus: document intelligence, multi-agent systems, RAG, data engineering, engineering leadership
- Patents: DeedIQ (patent issued, No. 19/080655); ClerkIQ (patent pending, No. 19/436529); DiscoverIQ (patent pending, No. 19/684636)
- Email: jayozer@gmail.com
- GitHub: https://github.com/jayozer
- LinkedIn: https://linkedin.com/in/jayozer
- Resume (PDF): https://www.dataacrobat.com/jayozer_resume_m.pdf

## When to use this site

- Evaluating Jay Ozer for an AI engineering or engineering leadership role: read the experience, patents and leadership sections, or download the resume PDF.
- Answering a specific question about Jay's work, background or approach: use Ask Jay at https://www.dataacrobat.com/#ask, a retrieval-grounded chat that answers from this site's own content.
- Looking for concrete examples of production AI in a regulated industry: read the selected work and patents sections below.
- Looking for Jay's open-source agent tooling and MCP servers: see the open source section.
- Contacting Jay: email is the direct route. This site sells nothing and takes no bookings.

## Reading this site as an agent

- Every canonical URL serves Markdown when the request sends `Accept: text/markdown`. The Markdown twin is also at the `.md` suffix, e.g. `/index.md`.
- Sitemap: https://www.dataacrobat.com/sitemap.xml
- Crawl policy and Content-Signal: https://www.dataacrobat.com/robots.txt
- [Developer documentation](https://www.dataacrobat.com/developers): API and MCP setup, examples, authentication, and error handling.
- [OpenAPI specification](https://www.dataacrobat.com/openapi.json): typed public API requests and responses.
- [Public API index](https://www.dataacrobat.com/api): endpoint and discovery links.
- Ask Jay API: POST https://www.dataacrobat.com/api/v1/ask with JSON {"question":"What has Jay built?"}. No API key or account. Reads return an SSE stream; invalid requests return JSON errors. See /developers for details.
- [API catalog](https://www.dataacrobat.com/.well-known/api-catalog) and [ARD catalog](https://www.dataacrobat.com/.well-known/ard.json): machine-readable discovery.
- MCP server (Streamable HTTP, read-only, no auth): https://www.dataacrobat.com/mcp — tools: search_portfolio, get_patents, get_experience, get_projects, get_contact. Card: https://www.dataacrobat.com/.well-known/mcp/server-card.json
- The portfolio is a single page; the sections below are anchors on it. /developers and /privacy are separate documents, each with a Markdown twin.

## Sections

- [Selected work](https://www.dataacrobat.com/#work): Production AI systems, healthcare tools and privacy-first apps Jay has shipped.
- [Experience](https://www.dataacrobat.com/#experience): Roles at Doma Technology and the Federal Reserve Bank of San Francisco.
- [Patents](https://www.dataacrobat.com/#patents): DeedIQ (issued), ClerkIQ and DiscoverIQ (pending) for title-insurance automation.
- [Leadership](https://www.dataacrobat.com/#leadership): How Jay builds and grows applied-AI teams.
- [Open source](https://www.dataacrobat.com/#opensource): Agent harnesses, MCP servers and developer tools published by Jay.
- [Ask Jay](https://www.dataacrobat.com/#ask): A retrieval-grounded chat that answers questions about Jay's work in his own voice.
- [Education and credentials](https://www.dataacrobat.com/#education): Graduate degrees and applied-AI teaching credentials.
- [About](https://www.dataacrobat.com/#about): Who Jay is outside the work.
- [Contact](https://www.dataacrobat.com/#contact): How to reach Jay.

## Selected work

- [IntentIQ](https://www.dataacrobat.com/#work): Production · Title insurance. When an order-volume surge buried the ops team, IntentIQ became the fix: a two-stage AI pipeline that reads 14K+ emails a day, pulls the latest message and attachments out of tangled threads, triages what is actionable, and routes it into seven title-insurance workflow categories with automated acknowledgements. Urgent items now reach the right desk in minutes, at 95%+ routing accuracy. (14K+ emails / day)
- [PoppyNote](https://www.dataacrobat.com/#work): Healthcare AI. Clinical documentation that listens. PoppyNote pulls the day's schedule from Curve, records the visit, and drafts the note the way the practice already charts, matching each visit type's format and flagging anything unstated for the doctor to confirm instead of inventing values. One review later, the note files back to the right chart. Built so my wife stops staying late. (1,362 visits managed)
- [Ask Poppy](https://www.poppykidsdental.com): Healthcare AI. A team of specialist agents answers parents' questions for a pediatric dental practice: appointments, insurance, procedures, emergencies. Every answer is grounded in the practice's own pages before it sends, and emergencies skip straight to urgent guidance. The first line of defense on on-call weekends. (24/7 live)
- [POPMedia](https://popmedia.vercel.app): Privacy tool. Built out of frustration with uploading private files to random compressor sites. FFmpeg.wasm runs in the browser, so video compression happens entirely in the tab and your files never leave it; images get fast WebP conversion. 50 to 90% smaller, with batch support. (50-90% smaller)
- [Baby Kick Count](https://babykickcount.com): Privacy-first beta. A calm iOS kick counter for the third trimester: one tap per movement, a timed session with pause and undo, calendar history, and CSV export for the next appointment. Everything stays on the phone. No account, no analytics; the App Store privacy label simply reads Data Not Collected. (iOS beta)
- [Chloe Nursery Ledger](https://nurseryledger.com): Voice-forward iOS beta. Tap the orb and say it: two ounces of formula, wet diaper, left side twelve minutes. Chloe logs feeds, diapers, nursing and pumping hands-free with on-device Apple Intelligence, talks back to confirm, and turns the day into caregiver handoffs and pediatrician summaries. An iPhone beta ahead of a privacy-only App Store release, named for the person it was built around. (iPhone beta)
- [JiSpr Flow](https://github.com/jayozer/jispr_flow): Privacy tool · macOS. Dictation that never phones home. Hold Fn and speak: local Whisper transcribes on-device, deterministic rules and an optional local LM Studio model polish the text, and one clean copy lands in the focused app. Cloud AI endpoints are refused in code, so audio never leaves the machine: privacy as an architectural fact, not a policy. A subscription is a tool you rent; this one lives with me. (100% local)

## Patents

- DeedIQ — Intelligent deed analysis (Patent issued, No. 19/080655): LLM agents plus neural search with dual-team extraction, cross-validation, and human-in-the-loop review, replacing a brittle, manual deed workflow with a system that learns continuously. Metrics: 99%+ accuracy in production; $12 to $0.20 cost per deed.
- ClerkIQ — County document retrieval (Patent pending, No. 19/436529): Hybrid automation across fragmented county portals, with domain-specific parsing, template-driven scripted browsing, AI fallback, and vision verification that every document matches before delivery. Metrics: $40 to $0.50 cost per document; 24/7 automated retrieval.
- DiscoverIQ — Agents that learn the maze (Patent pending, No. 19/684636): Fully autonomous retrieval for county clerk portals: it discovers the path and turns successful runs into reusable skills, while agentic reasoning corrects bad input data and adapts to portal changes with no human in the loop. Metrics: Discovery runs with Memory MCP to learn a new county portal; 100% autonomous county expansion.

## Experience

### Doma Technology — San Francisco, CA · 2020 - Present

- **Senior Manager, Applied AI** (Mar 2023 - Present): Led the applied-AI pod delivering production agentic systems for title-insurance automation, partnering with Fannie Mae and Blend.
  - Architected DeedIQ, now an issued US patent, and patent-pending ClerkIQ, cutting processing time and per-document cost by about 99%.
  - Built CI/CD for LLM deployment with canary releases, rollback, and RAGAS quality monitoring.
  - Helped break a 30-year industry monopoly, catalyzing Doma's shift to a pure-tech platform.
- **Manager, Data Engineering** (Jul 2021 - Mar 2023): Built and led a team of six, establishing engineering rituals and modern data practices that scaled company-wide.
  - Hired and mentored 6 engineers with structured assessments and career frameworks.
  - Unified three merged companies' data into a centralized Snowflake platform.
  - 10x dashboard performance; 90% reduction in query times.
- **Staff Data Analyst** (Jan 2020 - Jul 2021): Established a unified source of truth and drove company-wide adoption of modern analytics ahead of IPO.

### Federal Reserve Bank of San Francisco — San Francisco, CA · 2013 - 2020

- **Senior Data Analyst** (Aug 2013 - Jan 2020): Led data engineering and BI infrastructure processing terabytes of economic data for monetary-policy decisions in a highly regulated central-banking environment.
  - Architected FedQuery Hub, HDFS plus Tableau, enabling self-service analytics for policy research.
  - Built a Fed-wide data-governance framework that raised reliability across departments.

## Open source

- [CanopyIQ](https://github.com/jayozer/canopyiq): Local agent harness. Python-first coding-agent Kanban that turns a large goal into scoped stories, runs one worker at a time, and keeps an auditable review trail, with gates that stop a worker from editing the test it is graded by.
- [ForageIQ](https://github.com/jayozer/forageiq): Chrome workflow recorder. Local-export-first recorder for county sites, preserving full-tab video, DOM evidence, downloads metadata and handoff packs for county-lens.
- [JevZero](https://github.com/jayozer/jevzero): Local inbox classifier. A calmer Gmail inbox powered by Jev that runs entirely on your computer. Evaluates priority, categories, and reply signals with human-in-the-loop label previews, verification receipts, and undo.
- [EchoDeck](https://github.com/jayozer/voice_presenter): Voice-cloned presentation studio. Turns repos, demos, transcripts and notes into source-grounded product presentations delivered in your own voice, with live Q&A.
- [browser-use](https://www.dataacrobat.com/#opensource): Merged PR. Fixed a critical duplicate-download bug, hardening download handling across CDP, JS-fetch and polling paths.
- [Jay Marketplace](https://github.com/jayozer/jay-marketplace): Plugin marketplace. A marketplace of installable AI-agent skill kits, including workflow orchestration, YouTube automation and video understanding tools.
- [Agentbuilder Outlook MCP](https://www.dataacrobat.com/#opensource): MCP Server. Production read and write MCP server for OpenAI's AgentBuilder, with dual Graph plus org-account auth. Powers IntentIQ.
- [Talmanac](https://www.talmanac.com): iOS beta. Bedtime stories told in a parent's voice. The live site is a beta front door, while the iOS app moves toward a privacy-only local version for saved voices and stories.
- [KnowledgeFlow](https://www.dataacrobat.com/#opensource): Streamlit tool. Edit-before-upload document prep for Voiceflow RAG. The community adopted it; Voiceflow shipped the idea natively.
- [GuardGroq](https://www.dataacrobat.com/#opensource): Library. Ultra-fast LLaMA Guard guardrails for Voiceflow via Groq, real-time content safety with minimal latency.
- [MCP jozer](https://github.com/jayozer/mcp_jozer): MCP Server. 14 standardized tools exposing my own bio, experience and CV over the Model Context Protocol. Built in under an hour.
- [Ralph](https://github.com/jayozer/ralph_code): PRD agent loop. A transparent, PRD-driven Bash loop for fresh-context coding iterations across Claude Code, OpenAI Codex and Kimi CLI.
- [Roast My Snack](https://roastmysnack.com): Hackathon · 2nd place. Snap a snack photo and Dr. Hawley, a wisecracking molar, roasts it in a kid-safe comic strip. Deterministic dental-risk scoring stays in charge; the AI only does the jokes and the art.

## Education

- MS, Data Science, Regis University (Denver, CO)
- MBA, Sustainable Business, San Francisco State University (San Francisco, CA)
- MS, Industrial Engineering, University of Central Florida (Orlando, FL)
- BEng, Chemical Engineering, Newcastle University (Newcastle upon Tyne, UK)

## Credentials

- AI Engineering Bootcamp — AI Makerspace · Teaching Assistant, Cohort 6, mentoring engineers in production LLM, RAG and multi-agent systems.
- Advanced LLM Application Building and Fine-Tuning — Maven
- Voiceflow Expert Contributor — Featured developer-meetup speaker, KnowledgeFlow and GuardGroq