Ed L.

Most AI engineers ship demos that fall over in production. I design and build AI‑native operating layers that connect directly to real work: procurement, sales, finance, inventory, production, and quality. Recently I’ve been helping teams build agentic operating layers, like AOS, where different AI agents handle approvals, routing, data gathering, and decision support across multiple departments. Through this hands-on work, I have: • Designed agent harnesses and ML feedback loops so agents can be evaluated and improved over time. • Built RAG and LLM routing pipelines that choose the right tools and context for each request. • Developed vertical agents (niche, domain‑specific) with their own memory and reasoning layers, and full end‑to‑end UX. What I build • Agentic systems & multi‑agent orchestration – autonomous workflows, custom orchestrators, Claude/OpenAI agents, local + cloud LLM deployments. • Operational dashboards & internal tools – procurement, inventory, performance analytics, finance ops, trading ops. • Trading & market infrastructure – real‑time data pipelines, screening engines, execution terminals (live systems running with six‑figure capital). Recent production work • AOS – Multi‑agent operating layer coordinating procurement, inventory, finance, sales, and production for manufacturing SMEs. • LongShai / Dragon Screener – Real‑time AI stock screener for A‑share markets, integrated into live trading use. • BlackGator – Institutional BTC execution terminal with signal layer, wired into Polymarket and market‑data feeds. • Prism – Engineering performance analytics platform for an 80‑person IT team, shipped from blank page to production in two weeks. • BloomTechDaily – Fully automated AI newsroom: ingestion, filtering, summarisation, and publishing. How I work I don’t just write code or prompts. I partner with founders and operators to: • Map existing workflows in plain language. • Design the AI + automation architecture (agents, RAG, memory, evaluation harnesses). • Build, test, and deploy full‑stack systems that are observable, maintainable, and designed to survive real usage, not just demo day. Stack Claude, OpenAI, custom agent frameworks (graph‑style orchestration), RAG + vector search; React / Next.js / TypeScript, Python, Node; Postgres / Supabase; Vercel, Railway, GitHub Actions; Zapier / Make / n8n and API integrations when they’re the right tool. If you want an AI agent or operating layer around a specific part of your business (procurement, sales, trading, analytics, or internal ops), send me a message and we’ll design it together from workflow to production.