Building production-grade AI systems that capture real-world data, reason reliably, and automate workflows.
I’m an AI Engineer focused on building real-world, production-grade AI applications
I work across:
- Agentic AI + Multi-Agent Systems (tool-using workflows)
- RAG Pipelines (grounded answers with structured retrieval)
- Generative AI (LLM apps, prompting)
- Machine Learning (classical ML + deep learning)
A citizen-first platform designed to help people file complaints safely without going to the police station in the complaint filing process.
What it does
- Conversational AI chatbot that collects the right details step-by-step
- Persona-based personalization — questions adapt based on age, gender, and complaint type
- STT-powered citizen reporting (voice → structured complaint)
- Designed for secure reporting and reliable complaint capture
AI-powered bot rooted in Vaasthu Shastra principles.
- Uses RAG architecture with Qdrant Vector DB
- Delivers fast, accurate answers to Vaasthu-related queries
- Built for scalability and real-world usability
Tech: LangChain • Groq • Qdrant • Streamlit • Bolt AI
- Multi-agent orchestration (CrewAI, LangGraph)
- Tool calling, function execution, workflow automation
- Structured outputs, output parsing, validation-first design
- RAG: hybrid retrieval, contextual memory, grounded responses
- Vector DBs: Qdrant, ChromaDB, FAISS
- FastAPI services for AI products
- Dockerized deployments
- API design: auth patterns, reliability, scalability mindset
- Regression, Classification, Clustering
- ANN, CNN, RNN, LSTM
- Transformers & attention basics
- Text classification, NER, sentiment analysis
- Embeddings, tokenization, similarity search
- Transformer models: BERT, GPT, T5
| Project | Description | Tech Stack |
|---|---|---|
| Citizen Complaint Portal | Secure complaint system + AI-assisted structured intake for police workflows | FastAPI, Docker, LLMs, RAG, Vector DB |
| Vaasthu Vision AI | AI-based bot | LangChain, Groq, Qdrant, Shapely, Streamlit |
✔ End-to-End Builder — from AI pipeline → backend → deployable product
✔ Workflow-first AI — structured data capture + real automation
✔ Strong fundamentals — ML/DL/NLP + modern LLM systems
✔ High-stakes thinking — reliability matters more than hype
- Scaling agentic AI systems for real workflow automation
- Building RAG pipelines with better grounding + evaluation
- Improving AI observability using LangSmith + telemetry
“Build AI that ships — not AI that just sounds smart.”