I build production-ready AI systems that solve real-world problems, combining deep learning, NLP, and full-stack deployment expertise.
I'm an AI/ML Engineer passionate about delivering AI products that are practical, intelligent, and fast. My focus is on end-to-end solutions—from model development to scalable deployment.
- 🎓 Education: B.Tech in Computer Science & Engineering (2022–2026) @ Babu Banarasi Das Institute of Technology and Management, Lucknow
- 💻 Problem Solving: Solved 500+ problems on LeetCode (DSA)
- 🚀 Goal: Bridging the gap between state-of-the-art AI research and production applications.
| Domain | Technologies & Tools |
|---|---|
| Languages |    |
| Machine Learning |     |
| Generative AI & Search |     |
| LLMs & APIs |    |
| Tools & Deployment |     |
An automated travel planner that generates personalized itineraries by coordinating multiple real-time APIs.
🔴 GitHub • 🔴 Try Live Demo
- Tech Stack: Python, CrewAI, Gemini, Groq, FastAPI, Docker
- Developed an automated workflow using CrewAI and 6 AI agents to orchestrate live data extraction across flight, weather, and search APIs.
- Implemented LLM fallback across multiple models (Gemini → Groq) to maintain reliability during API rate limits.
- Deployed an asynchronous FastAPI backend to handle concurrent user requests in real time and generate Markdown travel guides.
A conversational AI pipeline allowing users to chat with their PDF data instantly.
🔴 GitHub • 🔴 Try Live Demo
- Tech Stack: Python, LangChain, Llama-3.1 (Groq), Gemini, FAISS, HuggingFace Embeddings, Streamlit, Docker
- Built a RAG pipeline using Llama-3.1 via Groq (primary) with a Gemini fallback, implementing optimized text chunking (512 tokens).
- Indexed a 10K+ document dataset using FAISS and local HuggingFace embeddings, achieving sub-200ms query responses on local testing.
- Containerized with Docker and deployed on Hugging Face Spaces for scalable hosting.
A semantic search engine allowing users to find clothing items using natural language descriptions.
🔴 GitHub • 🔴 Visit Live Site
- Tech Stack: Python, CLIP (ViT-B/32), FAISS IVF, TensorFlow (ResNet50), FastAPI, Docker
- Integrated CLIP (ViT-B/32) to map product images and text into a shared semantic space for accurate text-to-image matching.
- Improved search speed on a 44K product dataset from 2.3s to 180ms by replacing brute-force search with FAISS IVF indexing.
- Added a two-stage retrieval pipeline with cosine similarity re-ranking to improve final top recommendations.
A real-time music recommendation system that detects your mood via webcam.
- Tech Stack: Python, OpenCV, CNNs, FastAPI, Streamlit, Spotify API
- Trained a custom CNN for facial emotion classification and integrated the Spotify API for mood-matching playlists.
- Implemented Hybrid Collaborative & Content-Based Filtering to improve music personalization.