Asmit Kumar
Senior Backend & GenAI Engineer · Bengaluru
Backend engineer with 10+ years building systems that work at scale — data pipelines, distributed architectures, APIs, and now production AI.
Recently focused on building production GenAI systems — it's the backend fundamentals that make AI actually work in production. At Radar360, I build document intelligence platforms and LLM-powered workflows. Before that, I spent four years at Hearst Magazines scaling commerce data systems — 15M+ SKUs daily, AI-driven product matching that saved $200K/year, and a full Kubernetes migration of 60+ services. Started at Semantics3 as a data engineer, building web crawlers processing 10M pages a day.
Strong fundamentals. Practical AI. Production quality, always.
Skills
Backend / Infrastructure
GenAI / LLM
Experience
Senior Software Engineer
|
Dec 2024 – Present
Bengaluru
Document intelligence platform
Production — processing financial documents at scaleBuilt an AI-powered platform for automated document classification, structured data extraction, and conversational Q&A on financial documents. Integrated AWS Bedrock Claude with prompt caching for 80% cost reduction. Designed async FastAPI backend with hybrid PostgreSQL + S3 caching layer.
Multi-tenant analytics platform
Production — serving hundreds of organizationsContributed to the core backend platform handling dynamic query execution, pivot analysis, and async workflow orchestration. Database-per-tenant isolation with multi-database support (PostgreSQL, SQL Server, Oracle). Keycloak OAuth2 with PKCE, dependency injection, and OpenTelemetry tracing.
Dynamic report generator
Production — 40+ report types automatedDeveloped a backend service that dynamically generates professional Excel reports from JSON templates and live API data. Template-driven architecture with 59 configurations, dynamic formula generation, and cross-sheet cell reference tracking.
Senior Software Engineer
Dec 2020 – Dec 2024
Bengaluru
Product data-ingestion pipeline
20% decrease in time to create commerce articlesEngineered an end-to-end solution managing diverse data feeds via FTP, SFTP and direct APIs. Processed 15M+ SKUs daily, reducing reliance on crawling.
Product matching service
Saved $200K spent annually on a third-party solutionBuilt a product unification system across retailers using UPC and other universal product identifiers, combined with image and text similarity matching via OpenAI.
Product categorisation
Ad revenue up 8% due to improved targetingRe-architected the legacy categorisation service with a mapping system associating each product to its nearest node in the Google taxonomy.
Integrating fallback data sources
Improved data coverage and reliabilityDesigned a chain-of-fallbacks architecture for product data, routing through alternative data sources when primary feeds were unavailable. Custom fallback paths configurable per use case.
Infrastructure migration to Kubernetes
60+ microservices migrated with zero downtimeLed seamless transition of the full microservices fleet to K8s. Built custom data pipelines to maintain integrity during migration and sync 3TB of PostgreSQL data between AWS RDS instances.
Sales event optimization
Revenue impact of 10–15%Implemented Airflow pipelines for timely product updates and dynamic price adjustments during sales events.
Software Engineer
(Acquired by Hearst Magazines)Apr 2016 – Nov 2020
Bengaluru
Web crawler
10M pages/day across 300 e-commerce websitesCo-developed an in-house crawler with one main service dispatching tasks to 400 follower nodes across AWS, GCP, and third-party proxies. Implemented bot evasion, health checks, and efficient IP rotation.
Content extraction
Cost savings by eliminating crawls for sites with data issuesLed a team of 6 maintaining extraction rules for 1000+ e-commerce sites. Built an automated QA system to identify and flag data extraction issues.
Data Engineer
(Acquired by Hearst Magazines)Nov 2013 – Apr 2016
Bengaluru
Supervised data extraction & crawl validation
Reduction in customer complaints for data qualityAdded supervised extraction for sites based on requirements. Built crawl data validators incorporating schema validation, URL verification, and price validation.
Education
B.Tech, Computer Science Engineering
Punjab Technical University