Backend/Infrastructure Engineer focused on policy-based authZ systems (OPA/Envoy), ML inference infrastructure on GPU clusters, and distributed Go services. Currently building GPU-based inference services with gRPC and cloud providers(AWS/GCP).

India-based • Open to remote work with European teams

Gruve.ai

Software Engineer – 2 (Hybrid)| May 2025 – Present

• Designed and implemented an OTP-based authentication system by integrating Dex (OIDC provider) with Redis, extending the identity broker to support multi-factor authentication flows
• Engineered a token revocation mechanism by bridging Dex’s OIDC implementation with an internal RBAC manager, addressing missing functionality in the upstream open-source stack
• Architected the authentication and authorization domain (AuthX) in collaboration with principal engineers and the founding team, managing token lifecycle, session state, and RBAC enforcement across microservices
• Built and demonstrated end-to-end PoCs and internal demos, leveraging LLM-assisted development to accelerate API design, backend implementation, and test scaffolding under tight timelines
• Resolved a critical production bug causing platform-wide authentication failures, restoring service availability and preventing user lockout.
• Redesigned the authentication/authorization stack around OPA as sidecar policy decision point with Envoy ext_authz as the enforcement layer and Postgres RLS as a defense-in-depth backstop; separated policy (Rego, CI-built bundles) from live binding data served via a bundle server
• Designed EntitlementService as the policy administration/information point, owning RBAC binding writes independent of tenant lifecycle — decoupling tenant_id (billing scope) from org_id (authZ scope)
• Built JWT refresh-token architecture with jti-based rotation, token-family reuse detection, and family-wide revocation for gRPC clients, including proactive/reactive refresh strategies
• Led an Alluxio Enterprise AI POC as an S3 caching layer for vLLM model-weight loading on an H200 GPU cluster (8×H200/node, NVMe RAID5); root-caused etcd WAL fsync contention, JVM direct-buffer exhaustion under concurrent large reads, and TLS/zero-copy port conflicts in the S3 proxy
• Built a Go benchmarking harness (baseline, cache-miss, cache-hit, prefetch, concurrency) across 7B–70B model sizes, comparing RunAI Streamer against S3 vs. Alluxio-cached loads

Optum, UnitedHealth Group (Remote)| Software Engineer | Sep 2023 - Apr 2025

• Built Golang-ReactJS application replacing legacy IBM product, saving $80k in annual licensing costs
• Led security remediation initiative reducing 400+ critical/high vulnerabilities across production systems
• Migrated CI/CD infrastructure to GitHub Actions, halving deployment times while reducing infrastructure costs
• Developed authentication layer with OAuth2 and role-based access control for internal deployment tools

Optum, UnitedHealth Group (Remote)| Associate Software Engineer | Sep 2022 - Sep 2023

• Led enterprise migration from Team Foundation Version Control to GitHub Enterprise Cloud
• Rebuilt CI/CD pipelines in GitHub Actions, reducing Azure DevOps infrastructure dependency
• Automated workflow tasks using Linux scripting and GitHub REST API

State Street (Remote)| Software Engineering Intern | Jan 2022 - Jun 2022

• Built Golang CLI tools for workflow automation

My Journey in Software Engineering

I’ve spent the past few years building backend systems, automating infrastructure, and solving problems that matter. My journey started with a Go CLI tool during an internship at State Street, evolved through leading enterprise migrations and security remediation at Optum, and now focuses on architecting production-grade authZ platforms with OPA/Envoy and optimizing ML inference infrastructure on GPU clusters at Gruve.ai.

I’ve always believed in engineering systems that are both practical and robust — whether that’s replacing a legacy IBM product to save $80k/year in licensing, designing token-family rotations with reuse detection, or root-causing etcd WAL fsync contention on an H200 cluster.

Key Projects

Multi-tenant AuthZ Platform

Architected a production-grade authentication/authorization system for a multi-service gRPC platform: OPA as policy decision point, Envoy ext_authz enforcement, EntitlementService as the single writer of RBAC bindings, and Postgres RLS as backstop. Designed token-family-based refresh rotation with single-use semantics and reuse detection for revocation.

ML Weight-Caching Infrastructure (Alluxio + vLLM)

Operated and debugged an Alluxio Enterprise AI 3.8 deployment caching S3-hosted model weights for vLLM inference on H200 GPUs. Diagnosed etcd/worker I/O contention on shared RAID5, fixed TLS/zero-copy mutual exclusivity on the S3 proxy, and built topology-aware routing for cache locality.

Technical Skills

Languages & Frameworks: Golang • JavaScript/TypeScript • ReactJS • Python

Databases: PostgreSQL (RLS) • MySQL • MSSQL

DevOps & Cloud: AWS • Docker • Kubernetes • GitHub Actions • Azure DevOps • Linux Scripting • PowerShell • Bazel/Bzlmod

Architecture & Patterns: gRPC • REST APIs • Microservices • Distributed Systems • OPA/Rego • Envoy ext_authz • Alluxio • etcd • vLLM

Tools & Platforms: Git • GitHub API • NodeJS • Webpack • CI/CD Pipelines • Ubuntu

Focus Areas: Backend Systems • AuthZ Platforms • ML Inference Infrastructure • Performance Optimization • Cloud Migration • Security & Vulnerability Management

Let’s Connect

Open to backend, infrastructure, and ML-infra engineering roles focused on Go, Kubernetes, authZ platforms, and performance optimization.

Reach out via LinkedIn, Bluesky, or GitHub