Early engineer building Noah from the ground up. Full-stack ownership across React, Django, GCP, and LLM agent systems. Redefining coordination and communication as your primary intelligence layer

• Achieved near-peak hardware performance by optimizing FHE kernels on custom ASIC, accelerating encryption tasks by ~1800× over CPU.

• Achieved a 1.4× performance boost by engineering an out-of-order instruction compiler for custom hardware.

• Generated optimized microcode for relevant HE kernels such as NTTs, iNTTs, and Hybrid Key Switch.

• Helping build an LLM agent in Python + LangChain that monitors inboxes, classifies messages, runs class-specific data-extraction chains, and routes structured outputs to downstream systems.

• Improved LLM performance by testing 7+ LangChain flows in Python using OpenAI and Gemini API on held-out data.

• Reduced misclassification rate by 12% by enhancing RAG context with Chroma and adding HuggingFace reranking, retrieving higher-signal prior examples and user corrections.

• Designed and delivered NYU's K-12 Machine Learning program, adapting undergraduate-level curriculum into project-based modules for early learners.

• Taught 140+ students core ML concepts and applications, guiding projects from classic classifiers to simple generative models (GANs), with 98% successful completion.

• Introduced LLMs and agent-style workflows, teaching prompt engineering, RAG, and tool-using agents through guided projects and live demos.

• Improved dashboard performance by 25% by optimizing React visualizations and scaling support for larger datasets.

• Developed 12+ advanced features to track bot-like activity, coordinated sharing, and post virality.

• Cut API response latency by 28% by enhancing Django REST APIs and PostgreSQL schema design.

• Enhanced data interpretability by creating 4+ interactive data visualizations in React.