OPEN TO WORK • 📍 New York, NY • ✓ OPT Work Authorized
About Me
Building AI that works in production
I build AI that ships, and I care about the part after the demo. That means agents that finish the task, retrieval you can trace back to a source, and systems that hold up when real users show up.
Over the past 2 years I’ve worked across the AI stack, from data pipelines to multi-agent products: as the sole AI engineer on Bridgette at Bridgenow.ai, the sole backend engineer on MedEQ’s AI medical education platform, and on ML data pipelines at Cyware.
Outside work, I build fast. I’ve won the Vapi Voice AI and Self-Improving Agents hackathons in NYC. I’ve been the only AI engineer on a team more than once, so I’m comfortable owning everything from architecture and evals to APIs and getting it live.
Claude APILangGraphDSPyRAGFastAPIAzure OpenAIpgvectorDocker
Academic
Education
George Mason University
Master of Science in Computer Science (Machine Learning)
Jan 2024 – Dec 2025
GPA: 3.87
SRM University AP, India
Bachelor of Technology in Computer Science and Engineering, Machine Learning
Jun 2019 – May 2023
GPA: 3.6
Career
Work Experience
AI Engineer
MedEQ · USA
Apr 2026 – Present
Sole backend engineer on MedEQ, an AI-powered continuing medical education platform where clinicians practice against Claude-driven patient simulations; owned architecture across FastAPI, Cloudflare Workers, and Supabase (Postgres + pgvector).
Architected a 7-module CME accreditation system (identity & NPI verification, idempotent batch credit engine, typed/spoken reflection workflow, certificates, audit reporting) built to ACCME requirements with an external accreditation partner.
Rebuilt the FDA drug-label RAG ingestion pipeline with end-to-end source lineage across all 15 FDA label sections, so every clinical answer is traceable to its regulatory source.
Built a data-driven clinical simulation engine with severity-based patient vitals logic and orchestration observability across multi-call LLM flows, making AI behavior auditable for clinical reviewers.
Generative AI Engineer
Bridgenow.ai · USA
Oct 2025 – Apr 2026
Sole AI engineer on Bridgette, an agentic assistant for a B2B expert marketplace, replacing a 10+ step workflow (search → filter → invite → negotiate) with a single conversational interface; owned agent architecture, backend APIs, and frontend streaming.
Architected a production dual-agent system (companies vs. experts) using Claude and DSPy, with 8 Zod-validated tool schemas powering expert matching, engagement creation, and earnings analysis.
Raised end-to-end task success from ~62% to ~84% and cut hallucinated or incomplete responses by ~25% through validation layers and fallbacks, evaluated on 500+ test queries.
Reduced average response time from ~1.2s to under 400ms through prompt optimization and query-level caching.
AI Engineer
Cyware · Bengaluru, India
Jan 2023 – Dec 2023
Built Python pipelines to parse, normalize, and sanitize raw security logs into structured, high-signal datasets feeding downstream ML anomaly detection models.
Integrated frontend threat-intelligence dashboards with backend AI/ML services, resolving 150+ integration issues to deliver model-driven risk scores to SOC analysts in real time.
Led the Vue 2 → Vue 3 migration of threat-intelligence dashboards, improving rendering performance of live predictive risk-scoring views by 30%.
Improved product stability by 40% through systematic bug resolution in an Agile workflow (Git, Bitbucket, Jira).
Work
Featured Projects
Agentic AI
Guardian — AI Scam Shield for Families
🥈 2nd Place, Best Use of SurrealDB Agent Memory · Daytona HackSprint NYC 2026
An agentic AI system that intercepts scam texts sent to elderly parents, verifies them by calling the real family member, locks money requests behind a second family “key”, and explains the verdict to the parent in Telugu, English or Spanish.
Verify by callback: when a text claims to be family, Guardian calls the real relative on their saved number to confirm or deny it.
Money second key: money requests stay locked until a second family member approves with a press-and-hold. The AI can never approve money.
Safe link opening: suspicious links are opened in a throwaway Daytona sandbox via a cloud browser, analyzed, and destroyed.
Multilingual voice calls (Telugu, English, Spanish) with a family safe phrase and a teach-back question.
SurrealDB Agent Memory: every verdict is remembered, so repeat scams are caught instantly for any family, with a retrieval trace.
Rules-first safety: the LLM only extracts signals; a deterministic rules engine decides, and every warning cites the rule that fired. Live Arize evals on every case.
Full-stack AI platform matching patients to 2.8M+ providers in under 500ms using custom stream processing over a 704MB dataset, with 3D anatomy explorer and geolocation-based specialist ranking.
ML classification pipeline predicting 30-day hospital readmissions for diabetic patients trained on 100,000+ records, applying PCA for dimensionality reduction with measurable evaluation metrics.
Comparative ML pipeline evaluating KNN, Decision Tree, and Random Forest to forecast in-car temperature, achieving R² of 0.9997 with Random Forest using thermal imaging and augmentation for a hardware-free system.
10+ dynamic UI pages for the Andhra Pradesh State Warehousing Corporation government portal achieving 25% faster page load speeds, full mobile responsiveness, and improved backend data access.
Full-stack travel planning app integrating Google Maps, OpenWeather, and TripAdvisor APIs using React.js, handling 500+ monthly requests with reliable concurrent data fetching and client-side state management.
Complete UI/UX design and prototype for a modern carpooling application, focusing on intuitive user flows, scalable component architecture, and a clean mobile-first visual interface.
