Abhinav Patel

At Katonic AI, I operate observability infrastructure spanning Prometheus, Thanos, Tempo, and Loki, ensuring metrics, traces, and logs flow reliably across the platform. I also strengthen Kubernetes runtime security with Kyverno and Tetragon, and keep GPU training and inference workloads resilient during maintenance and upgrades.

I've deployed and maintained Kubernetes applications using Deployments, Services, Istio Gateways, and VirtualServices, while automating delivery with GitHub Actions and Argo CD. I improved Docker images by 35–50%, reduced reported vulnerabilities by 90%, and supported OpenShift certification by rebuilding images to run as non-root on Red Hat UBI.

At Hooman Digital LLP, I built reusable containerized templates for serving LLMs on Nosana. I added runtime model loading from HuggingFace through vLLM and Ollama, and resolved GPU compatibility issues that stabilized worker deployments.

I've also provisioned AWS infrastructure, designed isolated VPC networking, and configured load balancing, Auto Scaling Groups, and CloudWatch monitoring. My InferenceHub project brings together Kubernetes, vLLM, MinIO, Istio, cert-manager, and Prometheus/Grafana for secure GPU-accelerated LLM serving.