B.Tech CSE (AI/ML) · IIIT Nagpur · Class of 2026
Building things that turn LLMs into actual products.

I work at the intersection of LLMs, agents, and backend systems — building things where AI has to be reliable enough to ship, not just demoable. Most of what I build is one of two shapes:

• Agentic systems that plan → act → self-correct, not single-shot LLM calls
• Backend infra for AI workloads — routing, RAG pipelines, evals, observability

Languages    →  Python · Go · SQL · C++
LLM / AI     →  LangChain · LangGraph · Advanced RAG · Prompt Engineering
                Agentic AI · Vector DBs · Embeddings · Fine-tuning basics
ML tooling   →  Pandas · NumPy · Plotly · Hugging Face · Streamlit · Gradio
Infra        →  Docker · Kubernetes (Helm) · AWS (S3, Lambda) · gRPC / Protobuf
                Redis · Kafka · GitLab CI/CD
Observability →  LangSmith · Datadog · Grafana · Metabase
Providers    →  Groq · Gemini · OpenAI · Azure OpenAI · Bedrock

🤖 Autonomous Data Analyst Agent  ·  Live Demo ↗ Give it a CSV. It plans analyses, writes Pandas code, reads its own errors, self-corrects, and hands you insights — all in a live ReAct loop. LangGraph · Groq · Pandas · Plotly · LangSmith · Streamlit

Benchmarked head-to-head vs PandasAI on adversarial data: 8/10 data traps caught vs 6/10, 29% faster (141s vs 199s), 0 crashes.

🐛 Azure OpenAI Go SDK — Bug Fix Identified a routing flaw in the azopenai SDK affecting deployment-scoped routing for GPT-image models. Documented on Microsoft Q&A.

• 400+ DSA problems solved on LeetCode & CodeChef
• LeetCode peak rank 1575 · CodeChef peak rank 438 across 41 contests

• Agentic AI Certified Foundations Associate (Oracle)
• Andrew Ng's ML Specialization (Coursera)
• NVIDIA — Deep Learning Fundamentals
• NVIDIA — CUDA C/C++ Accelerated Computing
• Computer Vision · NLP

Open to conversations about AI infra, agent frameworks, RAG at scale, or LLM evals.