Leon Felipe S.

I build AI and optimization systems that make real engineering decisions (forecasting, dispatch, scheduling, planning), not just demos. I'm an electrical engineer and PhD candidate (defending in 2026) with 7+ years combining mathematical optimization, machine learning and power systems, plus industry experience in protection & control and grid operations. Top Rated Plus on Upwork. What I do: • AI engineering: LLM applications and agents, RAG, LLM-as-judge evaluation, FastAPI backends, async workers, Docker • Optimization: MILP/MINLP (Pyomo, HiGHS, Gurobi), genetic algorithms and hybrid MILP + metaheuristic methods for scheduling, dispatch and resource allocation • Forecasting & ML: multi-year demand/load forecasting with weather, demographic and technology-adoption drivers, including uncertainty ranges • Power systems & energy: EV / Vehicle-to-Grid integration, optimal power flow (pandapower, OpenDSS), BESS, microgrid protection coordination Recent work: • Load forecasting & V2G optimization for an electric utility (Upwork, NDA): long-term hourly load forecasts driven by EV adoption, weather and population scenarios, plus a genetic-algorithm V2G dispatch optimizer validated with optimal power flow on IEEE test feeders. Delivered as code, technical report and a journal-style manuscript. • AI voice interview platform (team of 6, top contributor): candidates complete a spoken interview with an AI interviewer; an LLM generates role-specific questions and scores each answer as a judge. My contributions centered on LLM integration and reliability: a configuration-driven provider layer (OpenAI / Gemini / deterministic mock for offline tests), asynchronous LLM evaluation on a Redis worker with race-free reconciliation, live speech-to-text transcript handling, watchdogs so speech synthesis can never stall a session, and secure single-use, expiring results links. • Food Opt (open source): multi-objective meal planner. Pyomo + HiGHS model that minimizes cost and prep time and maximizes variety under daily nutrition constraints, with a Next.js + Supabase app. If your problem needs AI that respects hard constraints (budgets, physics, schedules, capacity), send me the details and I'll reply with how I would model it.