At Gruve.ai, I built backend control plane microservices and a deployment API for model serving on KServe. The lightweight CLI replaced raw YAML manifests and Docker builds, reducing manual job configuration overhead by 60%.
At Rice University’s Transformational Technology & Innovation, I developed an AI-powered application and BERT-based NLP pipeline that classified support tickets with 88% accuracy and reduced triage time by 35%.
In my ClaimSense project, I built a distributed multimodal claims platform that processes audio, documents, and photos. I fine-tuned open-source LLMs and served models with vLLM, TensorRT, and ONNX, reducing inference latency by 65%.

