I am a Open Source Engineer specializing in Large Language Model (LLM) orchestration, RAG architectures, and scalable web systems.
My background blends frontier AI research with production-grade engineering. As a core contributor to the Google Gemini Cookbook (GSoC '25), I architected developer tooling and JS/TS workflows for the Gemini SDK. I currently focus on building high-performance, AI-integrated applications using modern full-stack technologies like Go, Next.js, and Supabase.
Core Competencies:
- Open Source Governance: Managing PR lifecycles and modernizing CI/CD workflows for large-scale repositories (Google DeepMind).
- AI System Design: Implementing hybrid RAG pipelines (Vector + Causal Graph) and optimizing inference costs.
- Full Stack Architecture: Building secure, concurrent microservices and interactive frontend applications.
| Domain | Stack |
|---|---|
| Languages | Python, Go, TypeScript/JavaScript, C++, SQL |
| AI & LLM Ops | Google Gemini SDK, RAG (Graph/Vector), LangChain, Ragas |
| Frontend & UI | React, Next.js, Tailwind CSS, Tiptap, Figma |
| Backend & Cloud | Node.js, PostgreSQL, Supabase, Firebase, Docker, Cloudflare Workers |
- google-gemini/cookbook: Active contributor/maintainer. Focused on scalable patterns for Gemini API implementation.
- GraphRAG-Analytics-Engine: Architected a hybrid retrieval system (Vector + Causal Graph) for analyzing high-volume unstructured customer support data.
- Architecture: Designed a dual-pipeline system comparing HNSW Vector Search (Qdrant) against Causal Graph traversal to solve multi-hop reasoning queries.
- Optimization: Engineered a parallelized metadata-enrichment pipeline using "Schema-First" processing, reducing data ingestion time by 92% (from 25 hours to 2 hours).
- Performance: Achieved 0.74 Faithfulness score (Ragas framework) using Graph-RAG, significantly outperforming standard Vector approaches by enforcing strict evidence-based citation.