At Ambee AI Incorp., I designed and implemented a multi-stage AI consultant backend that turns business problems into structured AI adoption strategies. I also developed a domain-aware reasoning engine that applies industry constraints, compliance filters, and risk assessment.
I built a tool-matching pipeline using sentence-transformers and cosine similarity, along with a scoring engine for ranking recommendations by relevance, cost efficiency, and deployment feasibility.
On my Digital Krishi Sathi project, I developed a multimodal farmer assistant that works offline and online. I quantized models for offline use, fine-tuned LLMs, and used FastAPI to orchestrate the workflow.
I also developed a RAG-based mail generator and a brain tumor detector. The detector used a CNN with ResNet transfer learning and achieved 98% testing accuracy and 99% validation accuracy on a dataset of 90K images.
