---
title: "Aashish Kumar - Software Engineer in Bengaluru | Tal"
description: "Meet Aashish Kumar, SWE @ Flash AI | IIT Kharagpur ’23 | Backend, Distributed Systems, AI Agents | Java · Go · Python · AWS based in Bengaluru. AI stack includes claude code, langgraph, and mastra."
image: "https://tal.club/meet/aashish-kumar-4/share-image"
---

# Aashish Kumar on Tal

## Headline

SWE @ Flash AI \| IIT Kharagpur ’23 \| Backend, Distributed Systems, AI Agents \| Java · Go · Python · AWS

## About

I build backend and infrastructure systems that move real product and cost metrics\. At Flash AI — used by 10M shoppers — I've driven \~$240K/year in cloud savings, built hybrid search over \~1M products at \~100K searches/day, and shipped the in\-app conversational shopping assistant\. Before that, at RootQuotient, I modernised a healthcare data platform for cancer\-registry workflows — moving legacy database logic into maintainable TypeScript services, building a reliable HL7 ingestion pipeline with distributed locking and worker\-level scaling, and raising backend quality through stronger test coverage and CI/CD\. I'm strongest in backend, distributed systems, infrastructure cost engineering, and AI agents at the application layer\. Core stack: Java, Go, Python, TypeScript, SQL, Spring Boot, FastAPI, Gin, Node\.js, PostgreSQL, Redis, Elasticsearch, Typesense, pgvector, AWS, Docker, OpenTelemetry\. I'm most interested in SDE\-2 / L4 / E4 / IC3 roles in backend, platform, and distributed systems, based in Bengaluru and open to relocation across India, the EU, or the UK\. If you're hiring for high\-ownership engineering roles where scale, system design, and business impact all matter, feel free to message me\.

## Location

Bengaluru

## AI stack

claude code

langgraph

mastra

openai

Google ADK

vLLM

## Work experience

Software Engineer at Flash AI, Dec 2024 \- Present · 1 yr 10 mos, Bengaluru, • Architected and shipped Flash AI's in\-app conversational shopping assistant end\-to\-end — a Google ADK multi\-agent backend with autonomous tool use, SSE\-streamed narration, and a prompt\-caching tier — scaling to 1\.4M monthly active users with a 20% retention lift\. • Designed and built a hybrid Typesense search layer over \~1M products, lists, and categories, now serving \~100K searches/day\. • Built Flash's catalogue\-ingestion pipeline with multi\-stage deduplication, reaching 95% merge accuracy across discovery surfaces\. • Re\-engineered email storage from the ground up — moved MIME bodies out of RDS into tiered S3 with lifecycle policies, re\-partitioned core tables with nested keys, and purged \~2B legacy records — cutting storage 66% and halving S3 write spend\. • Cut Flash's monthly cloud spend from \~$38K to \~$18K \(\~$240K/year\) by re\-architecting the AWS footprint: RDS rightsizing, ECS autoscaling, VPC subnet redesign, S3 Glacier tiering, and DB cluster consolidation\. • Led an organisation\-wide OpenTelemetry rollout across 28 services, migrating off CloudWatch to Last9 — cutting observability spend 30% while improving system visibility\.

Software Engineer at RootQuotient, Jun 2023 \- Nov 2024 · 1 yr 6 mos, Chennai, • Modernised a hospital cancer\-registry platform via a strangler\-fig migration from PL/SQL stored procedures and triggers to TypeScript services on SQL Server with Prisma — re\-platforming the REST API layer across 102 normalised tables in two phases\. • Replaced Oracle's time\-based DB scheduler with an HL7 ingestion pipeline running no\-overlap consumption on the cluster master — with database\-driven schedules, hold\-and\-retry semantics, and IPC\-propagated cron interval updates\. • Reimplemented the legacy \.NET SOAP webservices in Node\.js using WSDL XML contracts and edge\-js \.NET interop\. • Drove the test suite to 85% coverage, ran k6 load tests for on\-prem capacity planning, and owned the Azure DevOps CI/CD pipeline\.

Software Development Engineer Intern at Amazon, May 2022 \- Jul 2022 · 3 mos, Bengaluru, • Designed UUID\-based dashboard versioning on an EMR Spark job, lifting publish cadence from weekly to every second\. • Re\-architected an internal audit\-logging system: built a diff algorithm to surface row\-level changes between versions, and moved dashboard serving directly to S3, eliminating the need to maintain a million\-row DynamoDB table\.

## Education

Dual Degree \(Bachelor & Master of Technology\), Indian Institute of Technology, Kharagpur

[View the original profile](https://tal.club/meet/aashish-kumar-4)

```json
{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "Person",
      "@id": "https://tal.club/meet/aashish-kumar-4#person",
      "name": "Aashish Kumar",
      "url": "https://tal.club/meet/aashish-kumar-4",
      "image": "https://tal.club/meet/aashish-kumar-4/profile-photo.jpg",
      "description": "Meet Aashish Kumar, SWE @ Flash AI | IIT Kharagpur ’23 | Backend, Distributed Systems, AI Agents | Java · Go · Python · AWS based in Bengaluru. AI stack includes claude code, langgraph, and mastra.",
      "jobTitle": "Software Engineer",
      "worksFor": {
        "@type": "Organization",
        "name": "Flash AI"
      },
      "hasOccupation": {
        "@type": "Occupation",
        "name": "Software Engineer"
      },
      "homeLocation": {
        "@type": "Place",
        "name": "Bengaluru"
      },
      "alumniOf": [
        {
          "@type": "EducationalOrganization",
          "name": "Indian Institute of Technology, Kharagpur"
        }
      ],
      "knowsAbout": [
        "claude code",
        "langgraph",
        "mastra",
        "openai",
        "Google ADK",
        "vLLM"
      ],
      "mainEntityOfPage": {
        "@id": "https://tal.club/meet/aashish-kumar-4#webpage"
      }
    },
    {
      "@type": "ProfilePage",
      "@id": "https://tal.club/meet/aashish-kumar-4#webpage",
      "url": "https://tal.club/meet/aashish-kumar-4",
      "name": "Aashish Kumar on Tal",
      "description": "Meet Aashish Kumar, SWE @ Flash AI | IIT Kharagpur ’23 | Backend, Distributed Systems, AI Agents | Java · Go · Python · AWS based in Bengaluru. AI stack includes claude code, langgraph, and mastra.",
      "dateModified": "2026-09-05T07:57:24.332Z",
      "isPartOf": {
        "@id": "https://tal.club/#website"
      },
      "mainEntity": {
        "@id": "https://tal.club/meet/aashish-kumar-4#person"
      },
      "primaryImageOfPage": {
        "@type": "ImageObject",
        "url": "https://tal.club/meet/aashish-kumar-4/profile-photo.jpg",
        "contentUrl": "https://tal.club/meet/aashish-kumar-4/profile-photo.jpg",
        "caption": "Aashish Kumar profile photo on Tal"
      },
      "inLanguage": "en-IN"
    }
  ]
}
```