Experience
Roles in quantitative research, NLP for political-economy, robotics, and data analysis since 2022.
-
Built and deployed a FastMCP server that plugs into Claude Code to auto-reconstruct and edit generator offer curves on demand, cutting policy-revision time from hours to minutes.
- Built and deployed a FastMCP server that plugs into Claude Code to auto-reconstruct and edit generator offer curves on demand, cutting policy-revision time from hours to minutes.
TECHNOLOGIES
- FastMCP
- Claude Code
-
Developed an offer-policy generation tool using Markov decision models and Bayesian inference, transforming a previously manual process.
- Developed a cutting-edge tool that significantly enhances the accuracy of offer policy generation by implementing advanced statistical models for predictive analytics, leveraging Markov decision models and Bayesian inference, transforming a previously manual process handled by a team of employees.
- Collaborated with senior developers to refine algorithms, improving accuracy and efficiency of data-driven insights.
TECHNOLOGIES
- Markov decision models
- Bayesian inference
-
Engineered an NLP algorithm using HuggingFace and the GPT API to match historical tariff items with HS codes (~90% accuracy), and shipped a front-end that takes an image and outputs the matched HS code with confidence.
- Achieved a 90% accuracy rate in matching historical tariff items with HS codes by engineering a cutting-edge NLP algorithm using HuggingFace and the GPT API, and developed a front-end website to take in an image and output the corresponding HS code and its confidence.
- Increased prediction speed and precision by 50% by enhancing a machine learning model with parallel processing and caching.
TECHNOLOGIES
- HuggingFace
- GPT API
-
Closed the simulation-to-reality gap in robot behavior by training reinforcement-learning policies in PyTorch using OpenAI's PPO on Nvidia Isaac Gym and MIT SuperCloud.
- Closed the simulation-to-reality gap in robot behavior by innovating in reinforcement learning with Python and PyTorch, and training models using OpenAI's PPO algorithm.
- Showcased cloud infrastructure skills via Nvidia's Isaac Gym and MIT's SuperCloud by conducting advanced robotics research.
TECHNOLOGIES
- Python
- PyTorch
- OpenAI PPO
- Nvidia Isaac Gym
- MIT SuperCloud
-
Built forecasting models from historical data; presented insights to SOSV's CFO that supported a $225K capital raise.
- Catalyzed a successful $225K capital raise by presenting pivotal financial insights to SOSV's CFO, derived from robust forecasting and insightful research-led modeling of historical data.
- Secured company milestones by orchestrating data analysis and crafting key models for strategic decision-making.
TECHNOLOGIES
- forecasting
- predictive modeling
See Resume (PDF) ↗ for full detail.