Ganish Deepak N

Infrastructure & Platform Engineer @ Swiggy

2x SIH National Winner | 1 Granted Patent (202241060687)

Infrastructure & Platform Engineer with 2+ years of experience operating production Kubernetes, AWS, Terraform, and CI/CD platforms. Own and lead organization-wide AI/LLM infrastructure while building developer platforms as products for internal engineering teams. Experienced across production reliability, incident response, developer tooling, autoscaling, cost optimization, security, and AI-assisted operations.

Full Stack Development

ReactReact NativeNode.jsMongo DBPostgreSQLRedisTailwind

Backend Systems

Node.jsFlaskFast APIGraphQLRESTgRPCPostman

AI & Data Engineering

PythonC++vLLMllm-dLlama.cppLangChainPydanticHuggingFaceMCP

DevOps & Infrastructure

AWSTerraformDockerKubernetesIstioEnvoyKarpenterPrometheusGrafanaLoki

EXPERIENCE

Swiggy

Bengaluru, India

SDE 2 — AI Infrastructure & Platform Engineering

May 2026 - Present
Full-timeOn-site
  • Own and lead organization-wide AI/LLM infrastructure spanning the inference platform, LLM gateway, and internal MCP tooling.
  • Built and deployed an autonomous LangChain-based agent for independent production outage detection and mitigation, alongside a self-service MCP debugging server exposing 270+ Kubernetes tools — reducing mean time to first actionable debug insight from 24 hours to under 2 minutes.
  • Built and operated a production LLM inference stack on EKS using vLLM, llm-d, Karpenter, KEDA, and Istio, achieving 48.6 tok/s on Qwen3.8-27B in L40s with speculative decoding; published the architecture and operational runbook as open source.

SDE 1 — AI Infrastructure & Platform Engineering

Apr 2025 - Apr 2026
Full-timeHybrid
  • Owned a new scheduler trigger for Shuttle, Swiggy's internal CI/CD platform, across 4 repositories, enabling teams to schedule jobs declaratively through app.yaml without Terraform or Lambda; added 36 unit tests and drove UAT through rollout.
  • Led multi-cloud security incident response across AWS, Azure, and GCP following an exposed-credential and supply-chain incident; rotated 50+ compromised keys and coordinated organization-wide remediation end-to-end within a single weekend.
  • Led an organization-wide OpenSearch FGAC access-control migration using Terraform and IAM; authored the canonical reference implementation and rollout SOP, closing a cluster-wide authentication gap.
  • Designed and drove a zero-downtime OpenSearch cross-region replication (CCR) strategy, achieving 100% replication success with 1.52s average lag, and mentored two contractors through the rollout.
  • Migrated a single-instance production service to highly available ECS Fargate with canary rollout via Route 53, reducing infrastructure cost by 48.6%.
  • Built organization-wide CPU right-sizing automation across production Kubernetes clusters, saving approximately $3,000/day, and converted it into an ongoing Terraform-managed infrastructure pipeline.

Software Development Intern

May 2024 - Mar 2025
InternshipOn-site
  • Built a custom Prometheus exporter for Supervisord in Go and open-sourced the project.
  • Designed and built a distributed log-processing system — LogrCtx — using RabbitMQ/NATS Streaming for message routing across worker and drainer pods on Kubernetes, reducing ~900K log lines to 730 (99.92% reduction) in <2mins via RAG-based analysis.

EDUCATION

Bannari Amman Institute of Technology

Bachelor of Engineering, Computer Science and Engineering

2021 - 2025
  • 2× Smart India Hackathon National Winner
  • 1 Granted Patent (202241060687)

NOTABLE PROJECTS

LogrCtx

AI-Powered Log Analysis & Reduction

AI-powered log analysis tool reducing 900k logs to 730 (99.92% reduction). Features RAG-based insights, comparative analysis, and high-volume stream processing via SSE.

PythonFastAPIRAGLokiRabbitMQNATS

Inference Infra

In-house LLM Inference on Kubernetes

Production in-house LLM inference on EKS: GPU nodes provisioned on demand via Karpenter, vLLM served under llm-d, event-based autoscaling on queue depth with KEDA, and a hardened Bifrost gateway with hosted fallback.

KubernetesEKSvLLMllm-dKarpenterKEDA

Smart Whitecane

AI-Powered Assistive Device (SIH 2022 Winner)

Features realtime object detection, Google Assistant integration, and GPS tracking. Patented: 202241060687.

IoTAI/MLPython

OneBIT

Campus Utility App

Unified campus application for study materials and administrative tasks.

React NativeFirebaseEducation

LLM4SQL

Custom RAG implementation in DBMS

Functional MVP using React, Node.js, and Ollama. Custom RAG implementation in DBMS system for perfect query generation. Over 30+ stars and 5+ forks on GitHub.

FastAPILLMOpsFull stackRDBMS

Supervisord Exporter

Prometheus Exporter for Supervisord

A simple and efficient Prometheus exporter for Supervisord.

GoPrometheusDevOps

Nebula Dashboard

Mutual Funds Portfolio Tracker

Designed and built a dashboard tracking 45.1k live schemes.

MERN StackFinanceData Visualization

YouTube Automation

Content Management System

Automates content creation pipeline: Instagram login, reel downloads, compilation, and YouTube upload.

PythonAutomationYouTube API

HOBBIES

home lab serverspet projectsfreelancingguitar - beginnervideo games - pc/ps5