Applied AI & full-stack engineer building Python/TypeScript systems, LLM tooling, data pipelines, and developer automation

I'm Adam Owada, a Seattle-based Python and TypeScript engineer. I build practical software systems around LLMs, data pipelines, APIs, and developer workflows. My recent work spans Codex-native tooling, benchmark whitepapers, MCP integrations, sports-data scraping at multi-league scale, and Observe Safety, an alpha-stage B2B construction safety SaaS.

View projectsLinkedIn

Whitepapers

Published benchmark reports from my Codex subagent and LLM software engineering evaluation work.

First page thumbnail of the RuleLedger v3 white paper.

June 17, 2026

RuleLedger v3: Measuring Reasoning Effort in LLM Software Engineering

A 200-run benchmark showing high reasoning effort materially improved GPT-5.5 software-engineering quality, with xhigh producing the strongest observed tail behavior.

Open PDF
First page thumbnail of the Spark Mode Efficiency white paper.

June 25, 2026

Spark Mode Efficiency: Direct Edit vs Proposal Mode

A 360-row experiment showing direct-edit Spark subagents improved medium-level coordinators, while solo high/xhigh stayed ahead on peak quality and token efficiency.

Open PDF

Selected work

B2B SaaS

Observe Safety

Alpha-stage construction safety platform with Next.js, Expo React Native, FastAPI, PostgreSQL, tenant-scoped RBAC/RLS, analytics, mobile reporting workflows, CI, and AWS production-pilot infrastructure.

Agentic developer tooling

Codex Supervisor

Python-first control plane for Codex-driven engineering workflows, including durable task state, worker evidence, isolated worktrees, MCP/plugin surfaces, and review loops.

Python research systems

nlp-stock-prediction

Evidence-backed market research/reporting system with provenance, SQLite-backed artifacts, provider-health handling, MCP tooling, and evaluation workflows. It is not a trading app.

Paid client work

Agentify / Sports Business Technologies

Sports data ingestion and reconciliation platform work, including an 18-league scraper expansion with concurrent collection that cut multi-league runtime from roughly nine hours into the 3.5-4.3 hour range.

Work

  1. Observe Safety, LLC

    Co-Founder and Full-Stack Engineer

    Building an alpha-stage construction safety SaaS with Next.js, Expo React Native, FastAPI, PostgreSQL, tenant-scoped RBAC/RLS, analytics, mobile reporting, CI, and AWS production-pilot infrastructure.

  2. Self-employed

    Independent Applied AI / Full-Stack Engineer

    Selected independent software work across Codex-native developer tools, MCP integrations, Python research systems, published benchmark whitepapers, evaluation infrastructure, and full-stack product experiments.

  3. Sports Business Technologies / Agentify

    Contract Full-Stack Developer & Data Engineer

    Delivered sports data ingestion, reconciliation, salary-comparable analysis, and owner-facing admin workflows, then expanded the scraper toward 18 leagues with concurrent collection.

  4. Outlier.ai

    AI Model Evaluation Contributor

    Evaluated and rewrote model outputs for coding, function-calling, and tool-use tasks using detailed rubrics and quality criteria.

  5. Code Fellows, Inc.

    Lead Instructor, Python

    Taught advanced Python, Django, REST APIs, PostgreSQL, Docker, React/Next.js, data science, testing, and software engineering fundamentals.

Contact

Open to applied AI, full-stack, Python backend, developer tooling, and AI-assisted product engineering roles.

Email Adam