Building intelligent systems at the intersection of AI and software engineering
A builder at heart, passionate about making AI work in the real world.
I'm a UC Berkeley student pursuing a B.A. in Data Science and Computer Science (May 2027), passionate about building intelligent systems that solve real problems. My work spans AI research, full-stack engineering, and data science — always with a focus on production-grade quality and measurable impact.
At Berkeley's Sky Computing Lab, I architect multi-agent AI systems and design self-optimizing evaluation pipelines. My industry experience includes ML-based fraud detection at Autodesk, clinical dashboards at Artera AI, and full-stack platforms across multiple startups.
I thrive at the intersection of cutting-edge research and practical engineering — turning ideas into deployed systems that actually work.
A journey through research labs and high-growth companies, shipping real products.
Selected projects spanning ML, NLP, computer vision, and full-stack development.
End-to-end neuropharmacology drug discovery platform for designing, simulating, and analyzing novel ADHD therapeutics. Integrates AI-powered molecular design via Claude (Anthropic), real-time 3D visualization with 3Dmol.js, molecular docking & dynamics simulations, ADMET prediction, PubMed/ChEMBL database search, and team-based collaboration tools with regulatory pathway analysis.
Agentic operating system for biology researchers that orchestrates AI bioscience models (AlphaFold 3, ESMFold, RFdiffusion, DiffDock) into autonomous multi-agent pipelines. Researchers input a target sequence and BioOS handles everything: protein folding → binding site prediction → ligand docking → ADMET screening → FDA-grade documentation. Every step is reproducible, auditable, and compliance-ready.
Transformer-based recommendation engine with semantic vector search and LangChain orchestration. Features a Gradio dashboard with 90%+ recommendation precision, combining dense embeddings with collaborative signals for personalized results at scale.
RAG-based chatbot grounded in the Gale Encyclopedia of Medicine corpus. Combines Flask for serving, Pinecone for semantic vector retrieval, and a large language model backbone to deliver accurate, real-time medical Q&A with source attribution.
U-Net convolutional neural network for automated malaria detection from microscopic blood cell images. Achieved high sensitivity for parasitized cell identification. Research presented at the Society of Robotic Surgery Conference, demonstrating real clinical potential.
Tools and technologies I use to bring ideas to life.
Whether you have a question, opportunity, or just want to chat about AI — my inbox is always open.