At Flowdia, I co-founded the product and led engineering for an agent pipeline that generates production-ready full-stack apps from a prompt. I designed stateful iteration so generated apps can evolve while preserving schemas, UI state, and architecture decisions.
I built Flowdia’s infrastructure with Elixir/Phoenix, React, Nomad, Consul, and Traefik, and engineered model routing across multiple AI providers. Prompt caching and context pruning cut Developer-agent token costs across the pipeline.
At Accredible, I revamped an analytics service into Ruby lambda functions and migrated its storage to Hyperscale (Citus) on AWS EC2, using Terraform and Ansible. I also improved slow endpoints and queries, with response times more than three times faster.
Across roles at TaskRabbit, PassApp, Diatomz, and Verycreatives, I worked on backend and frontend products, infrastructure, and real-time features. At Bardy, I contributed to the backend, frontend, and infrastructure for a real-time feedback collaboration tool.
