Backend engineer building distributed control planes and infra automation in Go, Rust, and TypeScript. Recent work: a Kubernetes operator that collapses 9 manifests into 1 CRD, and an LLM-based CI failure triage tool. Currently building an autonomous DevOps agent for on-call incident response, and researching post-quantum authentication for Web3. Open to remote backend/infra roles and distributed systems consulting.
I'm a backend and distributed systems engineer focused on two things: infrastructure that heals itself and Web3 primitives that actually work at scale. I design systems where the control plane is cleanly separated from execution, where failure is expected and handled gracefully, and where observability isn't an afterthought.
Languages: Go, Rust, TypeScript Infra: Kubernetes, Docker, Kubebuilder / controller-runtime, GitHub Actions Data: pgvector, BullMQ Observability: Prometheus, Grafana
A Kubernetes operator that reconciles a single
AppServiceCRD into a full production stack — Deployment, Service, HPA, Ingress, PVC, PDB, NetworkPolicy, canary rollout, and Grafana dashboard — through an idempotent controller-runtime reconciliation loop.
Go Kubernetes Operator Kubebuilder controller-runtime CRDs GitOps Prometheus/Grafana
A sandboxed environment for running multiple LLM agents concurrently, coordinated through a shared vector-store memory (pgvector) and a task queue (BullMQ). Agents run in isolated containers with no direct host access. Built to explore multi-agent coordination patterns.
Autonomous Agents RAG + pgvector Gemini Function Calling BullMQ Multi-Process Architecture
A GitHub Actions-integrated tool that watches for CI/build failures, uses an LLM to do root-cause analysis on the failing logs, generates a candidate fix, and opens a pull request for human review. Cuts down manual triage time on CI failures — it doesn't auto-merge; every PR still needs human approval.
LLM Reasoning GitHub Actions CI/CD Automation Autonomous Agents
A Go-based reverse proxy that inspects incoming traffic for common attack signatures and routes suspicious requests to an isolated honeypot container instead of the real backend, with hot-swappable binaries for patching without downtime. Includes a Next.js dashboard for monitoring traffic and flagged requests, and uses an LLM (Cohere) to classify ambiguous payloads.
Go Next.js Cohere AI Honeypots Hot-swap Binaries
If you're working on something in distributed systems, Web3 infra, or autonomous agents — or you have a hard backend problem — reach out.
[email protected] · LinkedIn · Twitter @kritagyajha1 · Portfolio