Runs Optimization, AI, and Engineering at Electrotempo. Responsible for the full product lifecycle: formulation, implementation, and B2B delivery to ports, utilities, and fleet operators.

• Built and runs the team of scientists and engineers. The team's optimization and forecasting products generated $1.5MM+ ARR in the first year.
• Developed a fleet electrification optimization engine (MILP + decomposition). Went from prototype to production in five months, $500K ARR at launch.
• Built a joint site-fleet optimization platform that uses behavioral and scheduling data to minimize total cost of ownership. Cut lifecycle costs 12–31% for Port clients; models supported $320MM in infrastructure investment decisions.
• Built regional trucking power demand forecasting models using data fusion and regression. Three new utility contracts, $1MM ARR.
• Led the SOC 2 Type II compliance audit. Set up security policies, access controls, vendor management, and incident response from scratch to clear enterprise procurement.
• Product lead across all technical initiatives. Takes transportation and energy system constraints and turns them into shipped B2B optimization products.

Building an AI-powered supply chain optimization platform. Operations teams describe their problem in plain English; the platform formulates and solves the MILP. No OR consultant required.

• Built a natural-language-to-optimization compiler: a deterministic pipeline that parses LaTeX formulations into solvable MILP code. 83%+ accuracy on the NLP4LP benchmark.
• Wrote 23+ compiler subsystems covering parsing, semantic analysis, normalization, code emission, and programmatic constraint-sign validation. Replaced LLM-based validation with deterministic correctness checks.
• Built the full production stack: PostgreSQL (5-schema layout, Alembic migrations), RBAC auth (JWT, 4 roles, 16 permissions), Docker/CI, and a Compiler API with MCP support for AI-agent integration.
• Built a simulation scenario builder that prepares sensitivity analyses from solved models and uploaded datasets.

• Built a MILP model that automated short-term flow and carrier allocation across Amazon's US supply chain. Cut staffing costs 1.2% and dropped plan WAPE from 5% to 3.3%.
• The model enforces third-party carrier contracts and glide paths for peak event days. Reduced annual contract violation penalties by $500MM.
• Improved last-mile delivery topology optimization with predictive regression. Travel-time WAPE went from 16% to 11%.

• Improved post-disaster road recovery sequencing using dynamic programming and convex optimization, reducing solution complexity from factorial to exponential.
• Designed graph algorithms for oversize/overweight vehicle routing for TxDOT, cutting transportation and pavement costs by 23%.
• Won first prize nationally in the 2019 AWS & INFORMS Computing Cluster Competition with a joint demand forecasting/inventory control model.
• Developed efficient algorithms for symmetric traffic assignment, reducing solution time by ~50%, later adapting them for rail-road network electrification.