Paul Deng

Energy (Jul 2025 – Mar 2026); Vehicle (Mar 2026 – Present)

• Led Megacharger network planning to support Semi truck sales, designing siting and capacity strategy for 200+ locations and a multi-billion-dollar capital plan across North America, with rollout extending to Europe.
• Built point-based utilization forecasting in PyTorch to size and place new sites at arbitrary candidates, iterating across classical ML (gradient boosting, sequence models, neural ensembles) and LLM-driven autoresearch loops; currently internal, productionizing for customer-facing use.
• Extended firmware to enable trip planner support across new vehicle variants including Cybercab and Semi, contributing firmware code that integrates routing and charging logic for the new platforms.
• Architected and drove adoption of the team's LLM development harness, building and testing new skills, agents, and integration setups to deliver an order-of-magnitude lift in engineering throughput.

Energy · Charging Data Modelling

• Developed predictive models forecasting charging demand to guide expansion of the global Supercharger Network (4000+ sites) and Level-2 Wall Connector Network, totaling $500M+ in capital.
• Designed and maintained ETL pipelines on PySpark and Airflow, integrating multi-source data into KPIs and ML features.
• Ran statistical analysis and causal inference to inform monthly executive reviews on network performance and investment decisions.
• Partnered with infrastructure to deploy ML models on AWS Lambda and Kubernetes for production serving.

Advisory Services · Systems Analytics Group

• Built and calibrated travel demand models in Python, R, and EMME for government and transit agency infrastructure investments.
• Improved travel-mode forecasting accuracy with gradient boosting models over legacy logit baselines.

Civil Engineering, Transportation specialization

Thesis: A metaheuristic approach to vehicle routing problem with movement synchronization — genetic-algorithm-based metaheuristic for the traveling salesperson problem with drones.

Engineering Science, Infrastructure specialization