Full Stack AI Developer
Pune, Remote
Feb 2026 - Present
stealth
• Building full-stack client products with Next.js, TypeScript, and Node.js; REST and WebSocket APIs with real-time data flow and stateful session management between services • Integrating LLM APIs (OpenAI, Groq) into client apps chat interfaces, document Q&A, and tool-calling workflows with focus on minimizing time-to-first-token and perceived latency under production load • Shipped AI-powered features via streaming responses and async patterns; implemented function-calling chains with fallback handling for reliable agent behavior • Delivered a corporate website (Next.js + Node.js/Express, Vercel) with custom domain, SSL provisioning, and zero-downtime routing
AI Engineer Intern
India, Remote
Dec 2025 - Jan 2026
business xdata
• Architected a multi-agent AI system using CrewAI with modular pipeline design for automated UGC (image & video) generation, enabling independent scaling of each agent stage • Refactored FastAPI endpoints with async/await concurrency, cutting average response time from ~5,500 ms to ~500ms 90% latency reduction in production • Designed and implemented an end-to-end Retrieval-Augmented Generation (RAG) architecture using FAISS vector search and chunked document embeddings, enabling context-grounded LLM responses and reducing hallucination rate across agent outputs • Integrated LangSmith observability to track token consumption, API costs, and agent execution traces, surfacing prompt-level inefficiencies for cost reduction
Gen AI Intern
Remote, India
Jul 2025 - Aug 2025
growhut
• Built an LLM-powered scheduling agent using LangGraph integrated with GCP Mail and Calendar APIs, automating email triage and meeting scheduling end-to-end • Rebuilt AWS Lambda handlers for a real-time WebSocket audio streaming pipeline: redesigned connection pooling and async event handling to eliminate cold-start lag, reduce chunked audio ingestion delays, and improve end-of-utterance detection responsiveness under concurrent session load • Deployed Gen AI inference pipeline via Groq API achieving sub-100 ms time-to-first-token through structured prompt templates and output schema validation • Integrated Mem0 for persistent memory in a multi-turn chat app, enabling contextual continuity across long user sessions
Data Science Intern
Remote,India
Nov 2024 - Apr 2025
zummit infolabs
• Processed and cleaned 500K+ record datasets on Databricks, building reusable ETL pipelines for downstream ML training and feature engineering • Designed and deployed a collaborative filtering recommendation system (cosine similarity), improving personalization metrics; evaluated via MSE, MAE, precision@k • Built interactive Power BI dashboards to visualize model performance metrics and personalization KPIs, enabling stakeholders to track recommendation quality trends without querying raw data