Ask me directly
Research-grade AI, startup-speed operator
Ask about my background, role fit, Google/Nutrition5k, SnapCalorie, Parkour Fund, Hyperpilot, agents, or project tradeoffs.
Quin AI
Ask interview-style questions about my work, systems, and technical judgment.
Quin AI runs on a self-hosted inference stack with grounded retrieval over project docs, work history, and public technical context — a custom local RAG system designed to answer with more specificity than a generic API wrapper.
Quin Thames
I work across applied AI, data infrastructure, product systems, and operations — from first-author computer vision research at Google to production nutrition APIs, agentic QA loops, executive fund leadership, and trading intelligence platforms.
AI + computer vision
From Nutrition5k research to production food intelligence
I was first author on Nutrition5k at Google Research, designed the physical data-capture workflow behind the dataset, and later built production nutrition systems at SnapCalorie.
Production AI systems
APIs, agents, and workflows that actually ship
At SnapCalorie I built and own the Nutrition API for app/B2B analysis, built SEO/AI-search content systems, and led AI-agent adoption across engineering and operations.
Agentic software ops
Agents that test, file bugs, and feed fixes back into the product
In Hyperpilot, I set up automated agent accounts that actively use the platform, surface issues, file bugs, and hand work to coding agents for remediation.
Background
Build the data layer, ship the product, evaluate the system, automate the workflow.
AI + computer vision
I was first author on Nutrition5k at Google Research, designed the physical data-capture workflow behind the dataset, and later built production nutrition systems at SnapCalorie.
Production AI systems
At SnapCalorie I built and own the Nutrition API for app/B2B analysis, built SEO/AI-search content systems, and led AI-agent adoption across engineering and operations.
Agentic software ops
In Hyperpilot, I set up automated agent accounts that actively use the platform, surface issues, file bugs, and hand work to coding agents for remediation.
Data-heavy systems
Across JPL, DataPop, Parkour Fund, Google, and Hyperpilot, my work repeatedly turns large noisy datasets into decisions, maps, simulations, trades, or product workflows — often while also owning the operations around them.
Selected work
Applied AI, data infrastructure, agent workflows, mobile products, and market analysis.
AI-powered nutrition logging across photo, barcode, text, and recipe workflows, backed by production nutrition APIs, evaluation systems, and large-scale SEO surfaces.
Turned hard food-recognition and nutrition-estimation problems into production app, B2B, and growth systems.
Visit site →A trading and investing automation platform for analyzing ideas, running backtests, connecting to news feeds, and exploring AI-powered market workflows.
Brought strategy design, market data, news monitoring, agent QA loops, and automation into one workflow for faster investment research.
Visit app →A Google Research dataset and paper for automatic nutritional understanding of generic food, built around high-quality real-world capture and evaluation.
Created the research foundation for production food recognition and nutrition estimation work.
Served as CEO while building real-time on-chain analysis tools for crypto market movement, supporting analyst/trader workflows, leading the team, and running fund operations.
Combined executive ownership, data tooling, market analysis, team management, and accounting/operations in a live fund environment.
An iOS/React Native product exploring AI-assisted planning, brain dump capture, focus sessions, notifications, voice, and native app-blocking workflows.
Contract dev work for a small company — shows how AI can support structured product state and real user workflows beyond a simple chat box.
Contract dev work for a small company, currently in progress.
Smaller project, but a representative example of the kind of focused dev work small companies hire me for.
ISS imagery/geospatial matching and Mars Curiosity telemetry simulation.
Early proof of the same pattern: take messy high-volume signals and turn them into useful systems.
Google Search data analysis for advertiser insights and campaign performance.
Turning large-scale search behavior into actionable signal for advertisers.
Experience
AdWords/search diagnostics, Google Lens-adjacent visual recognition, Nutrition5k/CVPR.
SnapCalorie
Production nutrition API, AI food logging, website/SEO systems, agent workflows.
Hyperpilot
Trading automation, backtesting, news feeds, prices server, agentic QA loops.
Parkour Fund
CEO role spanning on-chain analysis tooling, analyst/trader workflows, team management, and financial/accounting operations.
JPL
ISS/geospatial image processing and Mars rover telemetry simulation.
DataPop
Google Search data analysis for advertiser insights.
Contact
Email me for roles, technical conversations, consulting, or a more detailed walkthrough of any project here.
Kauibungalow LLC · portfolio and project vehicle