AI/ML engineer building LLM systems, RAG pipelines and multimodal models — and shipping them with a real backend attached, not leaving them in a notebook that dies after the demo.

I care about evaluation-driven design: retrieval that actually retrieves, prompts that don't hallucinate, inference that stays fast under load.

• 🎓 B.Tech CSE @ Bennett University (2024–2028) · 8.65 CGPA
• 🧠 Working on — agentic workflows, semantic retrieval, multi-step reasoning
• 🥅 Building with — LangGraph, Groq, FastAPI, PostgreSQL, React
• ⚽ Off the keyboard — Subroto Cup captain, 3x Haryana state champion

AI Intern — AI Mill (Anudeep Ventures Pvt. Ltd.) · 06/2026 – Present

Shipping end-to-end AI products: system design, backend, LLM integration, testing, production deploy. Built and tested production solutions on LLMs, RAG, FastAPI, PostgreSQL and React, working across engineering teams.

                                                      🥅
                                               Docker  ·  Git  ·  Azure
                                                ── the last line back ──
                            
                                Python        FastAPI       PostgreSQL       JWT / RBAC
                                ───────── back four: nothing gets through here ─────────
                            
                                     LangChain      LangGraph      RAG + Chroma
                                     ────────── midfield: the engine room ──────────
                            
                                   React         Tailwind         Streamlit
                                   ───── front three: what the crowd sees ─────

LLMs & NLP
Transformers (BERT, GPT) · LangChain · LangGraph · RAG · Hugging Face · LSTM / GRU · Attention · Groq

ML & Data
Multimodal Learning · CLIP · XGBoost · Isolation Forest · EDA · Feature Engineering · Model Evaluation · Object Detection

Backend & Core
REST APIs · Pydantic · API Validation · Pagination & Filtering · Schema Design · SQL Aggregation · DSA (C++) · OOP

Agentic RAG system that generates syllabus-aligned exams from academic material and previous-year papers, then grades them. Retrieval, reasoning and generation are separated into modular stages so the outputs stay deterministic. A multi-step evaluation agent handles grading, feedback and weakness analysis.
Groq LangChain LangGraph RAG Chroma Firebase Streamlit

Autonomous AI Shopping Agents with Deterministic Guardrails, 4-Tier Hybrid Recommendations, Cryptographic Mandate Chains, and Merchant Autonomy Levers.
FastAPI SQLAlchemy SQL JWT RBAC

Predictive maintenance system — failure classification, Isolation Forest anomaly detection and rule-based risk scoring, with real-time predictions plus CSV batch inference through a React dashboard.
FastAPI Scikit-Learn Isolation Forest React Tailwind

93.7% accuracy on image–text mismatch detection across 700K+ cleaned samples. CLIP ViT-B/32 for joint image-text representation, XGBoost on top, augmentation pipeline for generalisation. Catches the ones that dive.
CLIP ViT-B/32 XGBoost NLP FastAPI React Vercel

Full platform for running eFootball tournaments — brackets, players, organizers, the lot. Football, but it's also the codebase.
TypeScript Full Stack Tournament Management

_{🪑 On the bench: Multimodal Emotion Detection — BERT text embeddings fused with MFCC audio features for emotion classification from conversation.}

⚽ Ship like you're chasing a 90th-minute equaliser.