I work on Computer Vision, Retrieval-Augmented Generation, Multimodal AI, and Applied Machine Learning Systems, with a strong interest in transforming research ideas into deployable solutions.
My recent work spans:
- 🔬 Visual Anomaly Detection Research
- 📄 Multimodal RAG Systems
- 🧠 Hybrid Retrieval Architectures
- 🏢 Enterprise Document Intelligence
- 🎓 Educational AI Systems
- 🤖 LLM-Powered Applications
Turning research into systems — and models into products.
- Multimodal RAG systems
- Enterprise document intelligence
- Computer Vision research
- Visual anomaly detection
- Educational AI applications
- Hybrid retrieval architectures
- AI engineering workflows
- 📝 Research paper under review at Springer Nature
- 🧠 Co-developed a lightweight anomaly detection framework
- ⚙️ Designed custom gating mechanisms
- 📈 Achieved 99.6% AUROC on MVTec AD
- 🚀 Reduced model size to 5.9M parameters
- 🔍 Combined global semantic consistency with local defect localization
Research Areas:
- Computer Vision
- Visual Anomaly Detection
- Representation Learning
- Lightweight Deep Learning Architectures
Dec 2025 – Mar 2026
- Developed secure Multimodal RAG systems
- Built hybrid semantic + keyword retrieval pipelines
- Integrated local LLMs for private AI workflows
- Designed offline document intelligence systems
- Worked on enterprise-scale document processing
Jun 2025 – Jul 2025
- Built internal AI chatbot using Flowise & Pinecone
- Developed LLM-based knowledge assistant
- Automated preprocessing workflows
- Created Power BI dashboards
- Performed data analytics and business insights generation
Django • XGBoost • REST API
- End-to-end ML pipeline
- Historical race analytics
- Live inference system
- Web deployment
🌐 https://f1predictor.onrender.com/
Streamlit • Gemini • SmartDataFrame
- Chat with CSV files
- Automated EDA
- Data insights
- Dynamic visualizations
🔗 https://github.com/MadeForMoney/Intelligent-CSV-Assistant-LLM-Powered
- Text + Tables + Images
- Hybrid retrieval
- Local LLM integration
- Website knowledge assistants
- FAISS & Pinecone
- Enterprise document intelligence
- Symptom extraction
- NLP classification
- Healthcare AI workflows
- Flask backend
- CBSE study assistant
- Personalized IIT/NEET learning system
- Adaptive explanations
- Retrieval-based learning
🥇 AIR 5 — Large Language Models (NPTEL)
🥈 AIR 16 — Responsible AI Systems (NPTEL)
📝 Research paper under review at Springer Nature
📈 99.6% AUROC on MVTec AD benchmark
💼 AI/ML Intern at Sopra Steria
💼 Data Science Intern at 8Queens
🎓 B.Tech AI & DS
📊 CGPA: 9/10
Research → Experiment → Failure Analysis → Iteration → Deployment
I care more about failure cases than leaderboard scores.
I trust ablations more than benchmark numbers.
I build systems first and optimize models second.
- Research internships (India & abroad)
- Computer Vision & Multimodal AI
- AI Engineering roles
- Enterprise AI systems
- Research publications
- Scalable ML systems
- Applied Generative AI
- I read architecture diagrams before conclusions.
- I trust learning curves more than accuracy scores.
- I enjoy reproducing papers before reading them fully.
- I care more about failure cases than perfect results.
- I’ve broken models intentionally just to understand them.
- I treat debugging sessions as experiments.
- My browser tabs usually contain papers, ablations, and benchmarks.
"Turning data into understanding — and models into systems."
⭐ If you've read this far, you might as well star a repository.