Software Engineer | AI Systems & Full-Stack Developer | Builder | Santa Clara, California | Ex-Capgemini | MS CS @ Santa Clara University

I build end-to-end intelligent systems — from scalable backend APIs to AI-powered applications used by real users.

• Applied AI (RAG, LLMs, multimodal systems)
• Backend & Distributed Systems
• Full-stack product development (web + mobile)

I care about shipping real products, not just prototypes.

• AI-powered systems (RAG, multimodal pipelines)
• Scalable backend systems (Flask + PostgreSQL)
• roduction-ready full-stack applications

Salus Wellness — Full-Stack Engineer

• Contributed to a production-grade wellness platform with web, mobile, and backend systems
• Built backend services for auth, scheduling, payments (Stripe), and messaging using Flask + Supabase
• Integrated AI-powered insights (GPT) for client recommendations and coaching workflows
• Worked on monorepo architecture (Turborepo) with shared types, APIs, and scalable infra

Sentari AI — Software Development Engineer (Full-Stack)

• Built & secured backend with 10+ Flask APIs (OAuth + JWT) for journaling, subscriptions & emotion tracking
• Integrated Whisper + GPT pipelines → improved transcription accuracy by 40%
• Shipped React Native iOS app serving 1K+ users

HCI Lab (SCU) — Graduate Research Assistant

• Built FocusMode Chrome extension used by 3,600+ users
• Designed AWS Lambda + DynamoDB backend handling 10K+ requests/day @ 99.9% uptime
• Improved dev velocity by 20% via CI/CD + monitoring

Capgemini — Software Engineer

• Worked on Android OS services for Logitech video conferencing devices
• Fixed race conditions & critical bugs, improving system stability
• Optimized OTA updates via S3 → 20% faster installs

• Built multi-tenant RAG system (Flask + pgvector + Gemini)
• Supports semantic search + citation-based QA over PDFs
• Designed scalable ingestion → embedding → retrieval pipeline

• Full-stack system (FastAPI + React) for emotional manipulation detection
• Combines rule-based + AI reasoning for risk classification
• Privacy-first (no data storage)

• Built multimodal pipeline using GPT-4 Vision + DALL·E 3
• Converts multiple images → structured prompt → generated composite
• Saves ~70% manual effort

• Streamlit + LangChain + ChromaDB + Gemini system
• Generates Q&A, tutorials, notes, practice questions from documents
• Uses local embeddings for cost efficiency

• Built TCP-based file transfer system with MD5 integrity checks
• Supports multi-file transfer over WiFi
• Faster than Bluetooth (~25%)

I enjoy building systems where AI meets real-world impact — from productivity tools to safety-focused applications.

Powered by coffee .....XD