I focus on building AI-powered systems and ML pipelines at scale.
I've architected recommendation engines for 30,000+ retail stores, optimized supply chains for Coca-Cola bottlers, engineered edge-optimized LLMs at Carnegie Mellon, And now, I advance business-critical analytical workflows using RAG systems.
I’ve dedicated my career to integrating AI into the product lifecycle and building the foundation that allows models to scale, adapt, and drive measurable business impact.
At my core, I enjoy building solutions that people rely on every day. If you’re working on something ambitious where my experience could help, I’d be glad to connect!
blah blah blah
Optimized inference pipelines for LLaMA2 and GPT-2 using speculative decoding, pruning, and quantization in PyTorch to reduce latency and memory usage. Built modular benchmarking workflows to evaluate accuracy–performance tradeoffs across CPU and GPU environments, targeting real-time and edge deployment scenarios.
Built an end-to-end ML system for traffic resource planning using near real-time data, focusing on reliable production use with MLOps. Engineered data pipelines, model evaluation, monitoring, and feedback loops required to sustain performance under changing real-world conditions. The resulting system maintained sub-5% forecast error while enabling continuous validation to support operational decision-making.
Designed a lightweight computer vision–based hazard detection system using deep learning in PyTorch, improving detection accuracy by 18% in critical regions. Optimized end-to-end inference and cloud-edge integration to achieve latency reduction, enabling deployment on edge devices for autonomous driving applications
Developed an end-to-end machine learning pipeline to evaluate consumer loan risk and optimize investment returns using historical LendingClub data. The project combined default prediction with return forecasting to move beyond binary risk assessment and toward decision-aware lending strategies. Using ensemble models and neural networks, the system achieved strong predictive performance (AUC 0.91) and informed a return-based investment strategy that increased profitability by 18% while reducing exposure to high-risk loans.
Building an AI-powered literacy platform that uses Large Language Models to turn reading into an interactive, gamified experience. Currently architecting the multi-tenant backend and implementing retrieval-augmented generation (RAG) to provide safe, age-appropriate content dynamically.
Industry, research, and applied ML work
Formal training and core coursework
Art + Writing outside of work
Led the creative direction of our university's largest media ord
ReadHead of Content for MIT's cultural fest hosting over 20k students
ReadWrote 4 Articles for the Editorial Board
ReadReach out for collaborations, roles, or a quick chat.