At Augai.ai, I operate the production LLM inference platform for codsy.ai, serving Qwen2.5-Coder-32B and DeepSeek-R1-Distill-7B on multi-GPU infrastructure. I also built a FastAPI router for distributing inference traffic across GPU workers and a custom parser plugin to resolve tool-calling failures.
I designed a hybrid GPU architecture with a bare-metal orchestrator and ephemeral RunPod compute workers. I’m the sole owner of the production infrastructure, including its Kubernetes cluster, developer containers, and monitoring.
At BoolMind, I architected APIs and microservices for Manage My Dojo, adding real-time notifications with Redis Pub/Sub and WebSockets. Delivering a robust beta contributed to a 25% increase in investor interest.
Earlier, at Techleadz, I built backend APIs and encryption services, and contributed to a Nextcloud file-management system. I began my career as a Machine Learning Intern at Technocolabs Softwares, contributing to supervised learning models and data-preprocessing pipelines.
