I'm an AI Engineer focused on building practical LLM, RAG, and Agentic AI systems that move beyond demos and actually work in production.
Currently, I work on scalable GenAI SaaS solutions, enterprise automation, RAG pipelines, and multi-agent workflows at UNIS / Item, where I help connect AI systems with real business operations, internal tools, and live enterprise APIs.
Outside of work, I’m building and exploring projects around MCP infrastructure, GraphRAG, Neo4j-powered tool graphs, retrieval-based agent routing, multimodal AI, and AI developer tooling.
I’m especially interested in one question:
How do we make AI agents more reliable, secure, and useful in real-world workflows?
You can also visit my website and ask my AI agent about my experience and projects:
Pratik Jadhav's Website
- LLM Engineering: Fine-tuning, PEFT, LoRA, QLoRA, quantization, model distillation
- RAG Systems: Self-RAG, GraphRAG, RAPTOR-style retrieval, hybrid search, re-ranking
- Agentic AI: LangGraph agents, MCP tools, multi-agent orchestration, dynamic tool retrieval
- AI Infrastructure: Docker, Kubernetes, MLflow, AWS SageMaker, Bedrock, API Gateway
- Applied AI: NLP, computer vision, multimodal AI, evaluations, automation workflows
A multi-tenant MCP gateway that helps AI agents securely access tools, reduce token usage, and choose the right tools with less noise.
A multimodal medical AI system that combines vision, clinical text, and LLM reasoning for healthcare-focused question answering.
Experiments around tool ranking, MCP orchestration, agent memory, and production-ready AI workflow infrastructure.
- I enjoy building AI systems from 0 → 1, especially where research meets real product use.
- I still love math, especially the kind that makes AI feel less like magic and more like structure.
- I believe the next wave of software will be built around agents, tools, memory, and context.
- I may or may not have too many cat memes saved on my laptop.
The best AI systems feel simple on the outside, but are deeply engineered underneath.