AI and data systems that repeat exactly, show their work, and ship. 15+ years turning messy data into production systems — from large-scale agricultural data pipelines at Bayer / The Climate Corporation, to Wear OS apps that pair straight into your own AI gateway — no phone, no cloud relay — to self-hosted stacks I build and run today, and teach you to deploy and own for yourself.
Seven lanes, one theme: systems whose outputs a peer can recheck line by line.
Deterministic agent architectures, LLM integration, and RAG/skill layers that repeat exactly and show their reasoning — not demo-ware. Also local inference: LLM, image, video, and audio generation tuned to your GPU, from consumer cards to workstation silicon.
Python · LLM · RAG · Local inferenceAg is home ground, not a side lane: research-trial pipelines, machine telemetry ingest (CAN → ISO-XML / ISO 11783), weather, imagery, and field-boundary fusion — with the crop-science background to validate the agronomy, not just move the bytes. 15+ years across Monsanto, The Climate Corporation, and Bayer digital-farm data teams. That's hard-won field and pipeline knowledge — not carried-out code or data — that My Farm Advisor is built on.
AgTech · ISO-XML · Research trialsKotlin-first mobile work, with a specialty in standalone Wear OS apps: no phone companion — the watch pairs with a 6-digit PIN over a private WireGuard tailnet and drives your agent in voice or text mode, with streamed UI and background services.
Kotlin · Wear OS · AndroidBatch and scripted ETL/ELT across raster, vector, tabular, and time-series data. The core strength: feature engineering that survives peer review — and model training, inference, and serving that scale without surprises.
Python · SQL · PySparkWeb, API, and product surfaces end to end — TypeScript front ends, Python back ends, data model and deploy included. One person from schema to shipped.
TypeScript · Python · APIsI don't just hand over a system and walk away — I teach you to run and extend it. Enablement is scoped to your stack and team: embedded sessions, code walkthroughs, and custom curriculum so you own what we built together. My ELVTR agricultural data course — farm-advisor skills plus deploying your own stack — is the public proof I can teach; private engagements cover any domain.
Workshops · Curriculum · FDE-styleYour stack on your hardware: Docker, Cloudflare Tunnel/Access, TLS, one-command deployments — and the documentation to hand it over.
Docker · Linux · CloudflareEverything I pitch, you can read line by line. Three repos I stand behind:
Python-first installer that stands up a full self-hosted stack on a fresh Ubuntu VPS: Dokploy, Cloudflare Tunnel, Tailscale, and LiteLLM — one AI key setup shared by every app. Ships pre-made Coder workspaces.
Standalone Wear OS client for the self-hosted Hermes AI gateway: 6-digit-PIN pairing, then voice/text chat with your agent over a private WireGuard tailnet — streamed markdown/mermaid, on-device TTS, run cancel, session resume, body tools. No phone app, no cloud.
Domain-specific skills for the My Farm Advisor agent ecosystem: portable, deterministic skill packs that turn farm, field, and trial data into analyses a peer can recheck line by line.
Fifteen years turning messy data into systems that ship — most recently as founder and instructor, before that inside one of agriculture's largest data platforms.
Independent consulting practice: deterministic agent pipelines and data engineering. Custom Python/Kotlin systems, self-hosted client stacks (Dokploy, Docker, Cloudflare), one-command deployments, and documentation-first handover.
Founded and operate My Farm Advisor — an agricultural data platform with a deterministic agent layer: portable skill packs that turn farm, field, and trial data into analyses a peer can recheck line by line. Full stack: Python/SQL ingest, geospatial pipelines, self-hosted Ubuntu deployments. myfarmadvisor.com ↗
Author and instructor of a hands-on course on building practical AI agents for agriculture: from data pipelines to LLM integration, taught to a professional audience on the ELVTR platform. View the course ↗
Built and operated large-scale ETL/ELT pipelines feeding weather, field, and imagery data into digital-farm products. Scaled feature generation across raw, derived, and engineered features; extended the model-validation framework to tie business metrics to model performance; designed geospatial ingest (CAN bus telemetry, satellite imagery, SSURGO) and self-service analytics tooling.
Verified details and working history also on my Upwork profile ↗.
Contract work, technical co-founder and engineering-lead conversations at small high-impact companies, strategic advising, architecture reviews, agent-pipeline rescue, or a course on doing this yourself — tell me what you're trying to build or own, and I'll tell you honestly whether I'm the right fit.
Also on LinkedIn ↗ — full background, ventures, and what I'm open to next.