I'm an AI Engineer at Trans Tech Projects Pvt. Ltd., Pune, working on Retrieval-Augmented Generation (RAG), multi-agent LLM systems, and production document intelligence pipelines. I hold a B.Sc. in Physics (Fergusson College) and an M.Sc. in Data Science (Symbiosis Institute of Geoinformatics).
At work, I've built:
- A Python + computer-vision automation pipeline that cut AutoCAD design turnaround time by 81%
- A FastAPI + OCR document intelligence system achieving 95% extraction accuracy on PDF-to-XML conversion at scale
I like building systems that actually ship — not just notebooks. Most of my projects below are deployed with CI/CD, containerized, and built around real production concerns like failover, observability, and cost.
paperbrain Production-grade RAG document intelligence system with a 7-agent pipeline (Extract → Analyze → Preprocess → Optimize → Synthesize → Validate → Assemble), FAISS vector search, real-time SSE streaming, and 5-provider LLM failover (OpenRouter → Groq → Gemini → HuggingFace → OpenAI).
hateguard-nlp Hate speech classifier with a 6-stage MLOps pipeline — LSTM model, AWS S3 storage, CircleCI CI/CD, and Docker + Flask deployment on AWS EC2.
tripmind-ai Multi-agent travel planning system (TaskflowAI) with 3 specialized agents, 4 real-time API integrations (Amadeus, Weather.com, Serper, Wikipedia), deployed on AWS EC2 with GitHub Actions CI/CD.
cotton-disease-prediction-deeplearning Deep learning project comparing multiple CNN architectures (including ResNet50) for cotton crop disease classification.
Languages: 
AI / ML: 
LLM APIs: 
Vector Stores / RAG: 
Backend: 
Data: 
Cloud / DevOps: 
- 🔍 Deepening production RAG architectures — retrieval quality, agent orchestration, and failover design
- 📦 Shipping document intelligence pipelines that hold up at scale, not just in demos
- 🤝 Open to AI Engineer / Data Engineer roles where I can own systems end-to-end
- 🧩 Exploring evaluation frameworks for multi-agent LLM pipelines
📫 Reach me at nachiketlohar0306@gmail.com or on LinkedIn