Samarth rawat

Designed and implemented the scalable core backend on Async Django, integrating structured logging and paginated APIs that dynamically query the Kubernetes API, improving customer experience. Tuned vLLM serving parameters for AI based inferencing, balancing Time to First Token (TTFT) and Tokens Per Second (TPS) based on client requirements. Extended Knative and KEDA on kubernetes, enabling autoscaling based on GPU metrics and scale-to-zero capabilities, which significantly reduced infrastructure costs. Authored a technical blog on Reducing GPU cold start time by over 70%, from 294 seconds down to 82 seconds, which increased organically driven traffic by 13%. Improved AWS CloudFormation templates for one-click provisioning of GPU-enabled EKS clusters, improving user onboarding experience. Served as a Forward Deployed Engineer, directly supporting clients to resolve production issues and optimize deployments. Technology Used - Django (Async), Kubernetes, AWS, CloudFormation, vLLM, redis, RQ.

Collaborated with CEO and CTO to secure $300k in funding in the first round, aligning technical capabilities with business objectives to attract investors. Led the integration of third-party platforms such as HubSpot, Slack, and others, enhancing the platform's functionality and streamlining user workflows. Managed end-to-end DevOps operations, including hosting servers and deploying large language models (LLMs), ensuring scalable, secure, and reliable infrastructure. Orchestrated Multi-Cluster deployment on Google Kubernetes Engine with rabbitmq and kafka configurations to improve scalability and deployment time. Implemented robust technical guidelines and comprehensive documentation to standardize development practices, ensuring scalable and maintainable codebases, improving team efficiency, and reducing onboarding time for new developers by 33%. Technology used - FastAPI, Modal platform, GitHub Actions, Flask, Azure VM, Kubernetes.