Atharva Patil

API Management & Gateway Platform

Go, Echo, Kong, Kafka, PostgreSQL, Kubernetes, Helm, Prometheus, Grafana

• Designed enterprise API platform handling 10M+ requests/day with 99.95% uptime — Kong gateway, Kafka event streaming, Go microservices on multi-node Kubernetes
• Implemented zero-downtime deployments (rolling updates, blue-green) and HPA handling 3x traffic spikes without latency degradation
• Built observability layer — Grafana dashboards over Prometheus metrics tracking p99 latency, error rates, throughput, and pod utilization with real-time alerting

Custom Kubernetes Job Controller

Python, Kubernetes CRDs, Informer Pattern, Helm, Jinja2

• Built custom controller watching multiple CRD types via informer pattern — auto-creates, rate-limits, and garbage-collects Jobs with exponential backoff retry
• Handled 100,000+ jobs with zero operator intervention; configurable concurrency limits and per-controller label scoping for namespace isolation

Code Compliance Platform

Python, FastAPI, Convoy, NATS JetStream, NoSQL, Kubernetes, AWS

• Built compliance platform scanning 100,000+ repositories — event-driven architecture with Convoy (webhook retries, DLQ) and NATS JetStream for guaranteed delivery
• FastAPI backend processing 200,000+ compliance checks/week; deployed on EKS with rolling updates and zero manual intervention
• Integrated AI-assisted code review workflows using Claude Code and custom automation pipelines

Unified LLM Gateway

Kong, Go, Lua, Kubernetes, Helm, Azure OpenAI, AWS Bedrock, VertexAI, vLLM

• Architected multi-provider LLM gateway (Azure OpenAI, AWS Bedrock, VertexAI) — custom Kong plugins in Lua/Go for intelligent routing, failover, and load balancing
• Handled 500K+ LLM requests/day with sub-200ms p99 routing latency; leveraged in-house API platform for auth, rate limiting, and quota enforcement enterprise-wide
• Self-hosted inference layer with vLLM reducing per-token cost by 60% vs cloud providers; reduced team onboarding from 2 weeks to 1 day

Enterprise Secret Scanner

Go, Gitleaks, TruffleHog, Django, NoSQL, Kubernetes

• Built secret scanning pipeline in Go (Gitleaks + TruffleHog) scanning 100,000+ repositories — identified 15,000+ secrets with 99.9% remediation within 30 days
• Integrated with custom K8s Job Controller for CRD-based scan scheduling; DefectDojo for enterprise-wide visibility, triage, and SLA tracking

Linux, AWS, Azure, Kubernetes, Docker, GitHub Actions, Ansible, ELK

• Provisioned and operated Kubernetes clusters and bare-metal infrastructure serving 72 teams; designed CI/CD pipelines reducing deployment time by 70%
• Implemented rolling updates, PDB, readiness/liveness probes for production workloads; automated infra provisioning with Ansible

• Fixed OAuth authorization server configuration for GitHub Enterprise Server (GHES) environments in Go
• Refactored API host resolver logic and enhanced test coverage across HTTP and OAuth modules