---
title: "Aravind Sundaresan - Founder / ML infrastructure Engineer in Chennai | Tal"
description: "Meet Aravind Sundaresan, AI Infrastructure Engineer | LLM Inference & Serving | Distributed Systems | vLLM Contributor | Microsoft, Ex-Amazon based in Chennai. AI stack includes antigravity, claude code, and hugging face. Explore public work links including github and anavindlabs.com."
image: "https://tal.club/meet/aravind-sundaresan/share-image"
---

# Aravind Sundaresan on Tal

## Headline

AI Infrastructure Engineer \| LLM Inference & Serving \| Distributed Systems \| vLLM Contributor \| Microsoft, Ex\-Amazon

## About

I build the platform layer between infrastructure and applications — control planes, scheduling, and developer tooling — most recently applied to ML serving and inference\. My core thesis: the bottleneck in AI isn't always the model; it's the runtime\. Backed by 7\+ years of distributed systems engineering at Microsoft and Amazon, I take infrastructure ideas from research → production → open source\. ⚙️ Production Infrastructure • Microsoft: Built distributed metadata infrastructure supporting 17,000\+ microservices, owning a 99\.99% availability SLA\. Focused on event\-driven systems, adaptive concurrency, observability, and developer tooling\. • Amazon: Engineered scheduling/orchestration for Alexa device infrastructure\. Built resilient fleet management with self\-healing automation, achieving 98\.4% scheduler uptime across 50\+ physical device categories\. 🚀 AI Infrastructure & LLM Serving • Adaptive Compute Orchestrator \(ACO\): Custom K8s scheduler extender on GKE \(Terraform, multi\-tenant isolation, Prometheus/Grafana\)\. Applies ant\-colony optimisation to cluster scheduling, improving placement efficiency by \~70\-80%\. • Clairvoyant: Drop\-in scheduling proxy for serial LLM backends \(Ollama, llama\.cpp\)\. Uses a lightweight ONNX predictor \(0\.029ms P99 overhead\) to reorder requests, reducing short\-request P50 latency by 70–76% under burst traffic\. \(arXiv:2606\.07248\) • vLLM Contributions: Two open PRs in vLLM's core scheduler, demonstrating hands\-on experience with production\-grade, parallelised LLM inference serving internals\. • ServiceScope v2: Agentic dependency mapping combining AST parsing with local LLM inference \(gemma3:4b\)\. Orchestrated with FastAPI, Celery, Redis, Postgres, Neo4j\. 🎯 What I'm Looking For: A role bridging research and production in AI infrastructure — building and operating platforms that make AI serving reliable, efficient, and scalable\. Focused on inference serving, scheduling systems, and multi\-tenant GPU platforms\. Core tech: Python • Go • C\+\+ • C\# • Kubernetes • FastAPI • Azure • AWS • Redis • Cosmos DB • PostgreSQL • Neo4j • ONNX • XGBoost • Prometheus • Grafana

## Location

Chennai

## AI stack

antigravity

claude code

hugging face

claudecode

Cohere

Synthesia

## Work experience

Founder / ML infrastructure Engineer at Anavind labs, Mar 2024 \- Present · 2 y 7 mo, Hyderabad, Independent Research Practice · Hyderabad, India • Independent systems research on LLM inference, GPU memory, disaggregated serving and training correctness; prototypes benchmarked on rented NVIDIA GPUs and GKE, results labelled measured, modeled or synthetic; published Clairvoyant \(arXiv:2606\.07248\); upstream vLLM and SGLang fixes

Software Engineer at Microsoft IDC, Aug 2021 \- Nov 2023 · 2 y 3 mo, Hyderabad, Service Inventory & Metadata Platform \(Internal Production Platform\) Designed and evolved Microsoft’s distributed metadata platform responsible for maintaining deployment truth across 17,000\+ microservices, enabling reliable service registration, validation, and recovery through event\-driven infrastructure\. Key Contributions • Architected a distributed validation platform using Azure Service Bus, Cosmos DB, Azure Functions, and Redis to process service metadata at enterprise scale\. • Designed adaptive concurrency control using Redis \(95% optimistic / 5% pessimistic locking\), eliminating cascading retry storms during production\-scale outages\. • Built Azure\-native active\-passive BCDR workflows with idempotent event processing, achieving zero\-data\-loss recovery objectives and sub\-minute failover\. • Developed Windows ARM64 Validation Provisioner using distributed worker nodes, reducing validation time from 60\+ minutes to \~20 minutes\. • Built a Test\-in\-Production telemetry framework using lightweight C\+\+ runtime instrumentation, adopted across 10 engineering groups to improve production observability\. • Automated Windows validation workflows using Power Automate and RPA, reducing UI\-related release defects by approximately 40%\. Technologies Azure Service Bus • Cosmos DB • Redis • Azure Functions • C\# • C\+\+ • Distributed Systems • Event\-driven Architecture • Observability

Software Engineer in Test Infrastructure at Amazon, Sep 2017 \- Aug 2021 · 4 yrs, Greater Chennai Area, Alexa Device Infrastructure Platform Designed distributed scheduling and orchestration infrastructure powering automated validation across Alexa hardware fleets\. Key Contributions • Engineered a deterministic scheduling platform orchestrating continuous workloads across 50\+ physical Alexa devices\. • Designed resilient fleet management through automated recovery, ADB keepalive monitoring, and self\-healing device orchestration\. • Eliminated non\-deterministic hardware failures through automated state management, Linux network isolation, and runtime health monitoring\. • Built Python\-based diagnostic tooling aggregating multi\-device crash logs, saving over 1,000 engineering hours annually\. • Automated firmware promotion and validation workflows across multiple Alexa hardware families\. Technologies Java • Python • AWS • Device Orchestration • Scheduling • Distributed Infrastructure • CI/CD • Linux

Technical Lead at OMG LABS, Apr 2017 \- Sep 2017 · 6 mos, Chennai Area, India, • Spearheaded customer acquisition through a subscription\-based model, boosting customer traffic by 15–20% in key market segments\. • Led end\-to\-end project delivery — from strategy and design to technical implementation and deployment\. • Managed vendor relationships and integrated e\-commerce tools such as site search and email marketing solutions, improving digital presence by 25–35%\. • Played a key role in achieving Amazon Launchpad selection, positioning OMG Labs among top emerging startups for international expansion\. • Coordinated across product, marketing, and tech teams to ensure agile execution and sustainable growth infrastructure\.

## Education

Bachelor of Technology \(B\.Tech\.\), Information Technology, Anna University Chennai

HSC, Computer Science, Velammal Matriculation Hr\.Sec\.School

## Public links

- github: https://github\.com/aravind0403
- anavindlabs\.com: https://www\.anavindlabs\.com/

[View the original profile](https://tal.club/meet/aravind-sundaresan)

```json
{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Person",
      "@id": "https://tal.club/meet/aravind-sundaresan#person",
      "name": "Aravind Sundaresan",
      "url": "https://tal.club/meet/aravind-sundaresan",
      "image": "https://tal.club/meet/aravind-sundaresan/profile-photo.jpg",
      "description": "Meet Aravind Sundaresan, AI Infrastructure Engineer | LLM Inference & Serving | Distributed Systems | vLLM Contributor | Microsoft, Ex-Amazon based in Chennai. AI stack includes antigravity, claude code, and hugging face. Explore public work links including github and anavindlabs.com.",
      "jobTitle": "Founder / ML infrastructure Engineer",
      "worksFor": {
        "@type": "Organization",
        "name": "Anavind labs"
      },
      "hasOccupation": {
        "@type": "Occupation",
        "name": "Founder / ML infrastructure Engineer"
      },
      "homeLocation": {
        "@type": "Place",
        "name": "Chennai"
      },
      "alumniOf": [
        {
          "@type": "EducationalOrganization",
          "name": "Anna University Chennai"
        },
        {
          "@type": "EducationalOrganization",
          "name": "Velammal Matriculation Hr.Sec.School"
        }
      ],
      "knowsAbout": [
        "antigravity",
        "claude code",
        "hugging face",
        "claudecode",
        "Cohere",
        "Synthesia"
      ],
      "sameAs": [
        "https://github.com/aravind0403",
        "https://www.anavindlabs.com/"
      ],
      "mainEntityOfPage": {
        "@id": "https://tal.club/meet/aravind-sundaresan#webpage"
      }
    },
    {
      "@type": "ProfilePage",
      "@id": "https://tal.club/meet/aravind-sundaresan#webpage",
      "url": "https://tal.club/meet/aravind-sundaresan",
      "name": "Aravind Sundaresan on Tal",
      "description": "Meet Aravind Sundaresan, AI Infrastructure Engineer | LLM Inference & Serving | Distributed Systems | vLLM Contributor | Microsoft, Ex-Amazon based in Chennai. AI stack includes antigravity, claude code, and hugging face. Explore public work links including github and anavindlabs.com.",
      "dateModified": "2026-10-29T10:44:31.232Z",
      "isPartOf": {
        "@id": "https://tal.club/#website"
      },
      "mainEntity": {
        "@id": "https://tal.club/meet/aravind-sundaresan#person"
      },
      "primaryImageOfPage": {
        "@type": "ImageObject",
        "url": "https://tal.club/meet/aravind-sundaresan/profile-photo.jpg",
        "contentUrl": "https://tal.club/meet/aravind-sundaresan/profile-photo.jpg",
        "caption": "Aravind Sundaresan profile photo on Tal"
      },
      "inLanguage": "en-IN"
    }
  ]
}
```