Open to opportunities

Biraj Mishra

@birajmishra

Message

Junior MLOps/DevOps Engineer building production CI, observability, and LLM pipelines end-to-end.

Nepal

Message

What I'm looking for

I’m looking for a team where I can own production MLOps/DevOps systems—CI/CD, observability, and LLM inference—while using staging, load testing, and infrastructure right-sizing to ship reliable improvements end-to-end.

I’m a Junior MLOps/DevOps Engineer with production experience across observability stacks, multi-tenant SaaS deployments, and LLM inference pipelines on Azure and Linux. I take services end to end—requirements and architecture, staging validation and load testing, resource right-sizing on constrained hardware, and ongoing production support.

Recently, I migrated production CI/CD onto a self-hosted GitHub Actions Runner Controller (ARC) farm on a bare-metal k3s setup with in-cluster caching and registry strategy, plus a staged release flow for reliability. On the AI side, I diagnosed queue contention on a shared-GPU Ollama inference server and tuned systemd settings to cut single-request latency to 0.10s (about 160x), while improving batch throughput around 7x.

Experience

Work history, roles, and key accomplishments

Current

Junior MLOps/DevOps Engineer

Current

Nepa Works

Nov 2025 - Present (8 months)

Designed and rolled out a Phase-1 staging deployment for a multi-tenant warehouse/inventory SaaS, implementing per-tenant PostgreSQL isolation with Prisma. Diagnosed and tuned a production GPU-based Ollama inference server to reduce single-request latency to 0.10s (~160x) while improving throughput.

Docker Compose GitHub Actions Wildcard DNS Routing PostgreSQL Prisma Ollama fastAPI Astro Nginx Azure Blob Storage K3s Docker In Cluster Registry Cache Ghcr.Io Fallback Load Testing Capacity Planning SonarQube

MLOps/DevOps Intern

Nepa Works

Nov 2025 - May 2026 (6 months)

Built an end-to-end observability and CI/CD foundation across production and staging, consolidating workflows and enabling real-time system visibility. Deployed production web apps with zero-downtime Nginx load balancing and configured MLOps tooling for orchestration, experiment tracking, and vector storage.