Agrim Sharma

At Surgeryview.ai with Capgemini, I lead backend architecture, cloud infrastructure, and ML deployment for a surgical-video AI platform while directing a 10-person engineering team. I re-engineered three legacy monoliths into async Python microservices on AWS ECS/Fargate, increasing throughput 40% and reducing downtime 25%.

I've deployed ResNet-50 and MSTCN inference pipelines on AWS SageMaker with BentoML and Terraform, reducing processing time from roughly 40 minutes to 6 minutes. I also shipped a RAG platform using LangChain, GPT-4, Gemini, and Hugging Face that serves 10K+ daily queries at sub-two-second P99 latency.

Previously, as the Python architect for AT&T at Capgemini, I decomposed a 400K-line monolith into 12 independently deployable microservices, lowered infrastructure costs 20%, and improved release velocity 25%. I built NL-to-SQL and asynchronous Kafka/Celery systems that enabled fast self-service analytics and high-throughput message processing.

Across 14+ years in backend engineering, I've led teams of up to 20, mentored engineers into senior roles, and delivered resilient systems with 99.9–99.99% uptime. I'm focused on practical GenAI, event-driven architecture, and cloud platforms that make engineering teams faster and products more useful.