Senior AI Engineer
I build production LLM systems, multi-agent architectures, and GPU-accelerated ML pipelines on AWS — currently leading ECLAT, a healthcare data platform for large-scale OCR and clinical terminology extraction.
Senior AI Engineer with 7+ years of software engineering experience, focused on production LLM systems, multi-agent architectures, and GPU-accelerated ML pipelines on AWS. Currently designing and building ECLAT at Maitri Holdings — a healthcare data platform combining large-scale OCR, clinical terminology extraction (LOINC / RxNorm), and orchestrated GPU inference. Strong full-stack and cloud-native foundation, with a proven track record leading cross-functional Agile teams and translating architecture decisions into measurable cost and latency improvements.
Lead engineer on ECLAT, a GPU-accelerated healthcare data pipeline covering document OCR, medical terminology extraction, and batch plus online model inference on AWS. Designed the GPU orchestration architecture — SageMaker Async Inference, AWS Batch, and Step Functions — with a tiered execution strategy (scale-to-zero for bulk backfill, warm fleets for latency-critical paths) that materially cut idle GPU spend. Built extraction services mapping unstructured clinical text to LOINC and RxNorm codes, and own end-to-end delivery: infrastructure-as-code, CI/CD, observability, and production reliability.
Led development and integration of cookie-based authentication across TaxMD's multi-agent system, covering every endpoint. Integrated AI-driven capabilities into tax analysis tooling, designed scalable backend services, and strengthened platform security with advanced validation protecting sensitive tax data.
Led a 10-person cross-functional team as Technical Lead across multiple AI-driven client projects. Spearheaded Google's Agent-to-Agent (A2A) architecture in a multi-tenant GenAI SaaS platform with Next.js and FastAPI, embedding Gemini 2.5 as the core LLM for real-time reasoning and autonomous task execution. Engineered secure REST APIs and WebSocket channels that drove a 40%+ increase in user engagement, and championed observability with Prometheus and Grafana.
Developed and deployed a Rasa-based chatbot system with Kubernetes Secrets for secure credential management. Implemented Velero for disaster recovery and Skaffold for infrastructure provisioning within Codehub, a platform enabling students to learn AI by running Python programs.
Led customization of JupyterHub and Jupyter Notebook environments for AI education. Revamped Codehub for 1,000+ active users — a 40% increase in engagement and 30% improvement in satisfaction — deployed on GCP with Kubernetes and Docker.
Developed and maintained fuse.ai, an AI learning platform for students, and applied machine learning techniques in the intern training programme — including hand recognition and OCR projects.
Team Lead
Directed web and mobile development on AWS with a Python backend, SNS notifications, and SAM CLI serverless deployment, serving 500+ daily active users. Drove Agile delivery to a 30% productivity gain and a 50% reduction in user-reported issues.
Food Coupon & HR Management Systems
Built end-to-end observability across web and mobile using Prometheus, Tempo, and the ELK stack, improving incident response time by 20%. Validated scalability with k6 load testing under high-traffic conditions.
Backend Developer
Built online examination, quiz, and grading features plus APIs for real-time student progress tracking and detailed reporting across multiple institutions.
Let's connect — whether you need production AI systems, cloud-native solutions, or just want to grab a coffee.