AI Engineer at DataCore Softwares

Built AI-powered storage services with Python, FastAPI, Docker and self-hosted LLMs for real-time detection, recognition, transcription, summarization, and translation. Engineered FastAPI RESTful APIs, containerized with Docker, and deployed on AWS EKS; achieved <200 ms inference and 99.9% uptime at peak load. Fine-tuned open-source LLMs (LLaMA 3, Mistral 7B) on custom data; optimized GPU inference via TensorRT + ECR images, cutting costs 12%. Built end-to-end ML pipelines: video frame extraction → segmentation → vector embedding → similarity search. Implemented CI/CD with GitHub Actions, Terraform IaC, and Prometheus/Grafana monitoring; automated rollbacks cut deployment failures by 60%.