I'm an MS Computer Science graduate from the University of Illinois Chicago (May 2026, GPA 3.89), building agentic AI systems, RAG pipelines, and full-stack products in Python, Go, and TypeScript. I'm on OPT and open to full-time AI Engineer, ML Engineer, and SDE roles at startups and mid-sized companies.

📍 New Jersey · ✉️ gaurav.pvt25@gmail.com · 🔗 LinkedIn · 🌐 Portfolio

• Shipping Winnow, a local-first AI inbox triage agent with a tiered classifier + LLM architecture
• Building autonomous multi-agent systems with CrewAI, PydanticAI, MCP, and LangGraph
• Contributing to open-source agentic frameworks (OpenSRE, exploring CrewAI and Letta next)

📬 Winnow · Python FastAPI PydanticAI Next.js Postgres scikit-learn

Local-first AI inbox triage with a tiered classifier + LLM architecture. A local scikit-learn + MiniLM classifier handles the majority of routing in ~5ms on CPU at $0; a PydanticAI agent runs only on uncertain cases. Nightly retraining loop learns from user corrections with regression guardrails. 153 passing tests, full-stack (FastAPI + Next.js 15).

🔧 Autonomous Agentic SRE Pipeline · Go Python MCP Kubernetes CrewAI Gemini

Self-healing pipeline that polls Prometheus, diagnoses incidents via an MCP-bridged LLM, and pushes strategic merge patches to resolve OOM errors without human intervention. Go watchdog triggers CrewAI agents that analyze pod logs in a closed remediation loop with escalation memory (restart first, patch limits second).

⚙️ ToolGen · Python OpenAI NetworkX Pydantic Multi-Agent LLM-as-Judge

Multi-agent synthetic data pipeline that generates multi-turn tool-use conversations grounded in ToolBench schemas. LLM-as-judge scores each conversation on naturalness, tool correctness, and task completion, with automatic repair loops on low-quality outputs and cross-conversation diversity steering for training-set variety.

📚 AI Research Assistant · Python LangChain FastAPI Pinecone

Production RAG API for querying academic documents in natural language. Sub-200ms p95 under concurrent load with 95% retrieval consistency. Automated regression testing via GitHub Actions.

🤖 AI Documentation Bot · Python LangChain Pinecone OpenAI

Semantic search across Pandas, NumPy, and Scikit-learn docs. MLP classifier ranks results by source reliability, improving answer relevance by 20% by prioritizing official API references over community content.

OpenSRE · ⭐ 1,900+ stars · AI-powered SRE agent framework

Redesigned the root-cause analysis prompt with a structured multi-step reasoning sequence (hypothesis generation, evidence validation, elimination). Added an alternative-hypothesis output section that reduced false-positive classifications by 30%, plus a backward-compatible response parser for mixed-signal incidents.

Languages Python · Go · TypeScript · JavaScript · Scala · SQL · C

AI / ML PyTorch · TensorFlow · LangChain · CrewAI · PydanticAI · MCP · DSPy · PEFT · Transformers · RAG · Pinecone · Sentence Transformers · Scikit-learn

Backend & Infra FastAPI · Next.js · Postgres · SQLAlchemy · Kubernetes · Docker · Prometheus · Spark · AWS · Azure · Vercel · Railway · GitHub Actions · Linux

Data & Viz Pandas · NumPy · OpenCV · Tableau · Streamlit

If you're hiring for AI/ML engineering, agentic systems, or full-stack ML products, I'd love to hear from you. The fastest way to reach me is gaurav.pvt25@gmail.com or a LinkedIn message.