How I got here.
I started on factory floors. At Tesla I built material flow models and warehouse layouts across three production shops, and identified millions in savings in dock allocation and storage. At Boeing I keep an eight-person engineering program on schedule as Scrum Master and built the data systems that cut executive reporting time by two thirds. Industrial engineering is where my AI work comes from. I think in inputs, constraints, and outcomes, and I build systems I can trust to run on their own.
In early 2026 I started building agentic AI on nights and weekends. It turned into Soltreya, a quality platform for regulated manufacturers that runs on a multi-agent framework I built from scratch. I'm now all in on AI engineering.
Where I've worked
- Architected the WAT framework (Workflows, Agents, Tools) so multi-step AI runs unattended without compounding errors: I scope Claude to the judgment calls and keep execution in deterministic Python I wrote.
- Run the company on it in production: ~4,000 scheduled and event-driven runs a month on Trigger.dev, a Claude orchestrator routing to 7 specialist agents, 40+ Python tools, and 17 background jobs.
- Built the product (Next.js, Supabase, Stripe) to a 100 Lighthouse score with zero accessibility violations, behind a four-gate security model (rate limiting, auth, tenant isolation, cost caps).
- Keep an 8-person engineering program across 3 partner companies on schedule as Scrum Master, surfacing interface risk early through technical syncs and milestone reviews.
- Cut executive briefing prep time 67% by designing the program's authoritative data source and automated KPI dashboards (Qlik Sense).
- Identified $3.4M in dock-allocation savings and cut storage 66%; redesigned delivery routes for $4M/yr additional OPEX reduction, freeing 16,000 sq ft.
- Built material-flow models for 3 shops (250+ parts) in Revit and AutoCAD; raised lineside part buy-off rate from 54% to 95%.
- Automated a hospital therapy department's daily productivity reporting, cutting supervisor reporting time 66% (90 to 30 minutes) and removing manual calculation errors.
What I've built
Soltreya: AI quality platform
Soltreya helps regulated manufacturers handle corrective actions: the root-cause and paperwork work quality engineers otherwise spend weeks on per finding. It finds and scores target factories, drafts outreach in my voice for me to approve, runs the live Quality Hub product, and keeps itself running across 17 background jobs behind a four-gate security model. I built and operate the whole thing myself, for manufacturers working under standards like IATF 16949 and ISO 9001.
The WAT framework
Workflows, Agents, Tools. The reliability pattern under everything: probabilistic AI handles reasoning while deterministic code handles execution, so 90%-accurate steps never compound into failure.
40+ Python tools
40+ Python tools and 17 TypeScript background jobs for sourcing, research, reporting, and outreach. They run on a schedule or fire from Telegram, email, and cron, and connect to 10+ external APIs.
What I work with
AI & Automation
Languages
Web & Infra
Systems & Process
Modeling & CAD
Domain
Foundation
Let's talk.
Questions about the WAT framework, Soltreya, or anything else on this page? Reach out. I read everything that lands in my inbox.