Headline

SDE-1 @Karbon Business

About

Experienced in backend systems that don’t just scale—they adapt. Working as an SDE-I at Karbon, I specialise in designing fault-tolerant, event-driven services and AI-integrated workflows that transform complex operations into reliable, self-recovering systems. My core focus is applying AI within production-grade architectures, ensuring reliability and measurability over “just model accuracy.” I work across FastAPI, Django, OpenAI SDK, and LangGraph to enable intelligent routing, recovery-based event handling, and high-performance API design. Outside work, I actively build AI-driven MVPs and SaaS products aimed at reducing operational friction and enabling faster decision-making. • Architected recovery-based event routing guaranteeing exactly-once processing and automated retries. • Built agentic workflows with LangGraph to enrich and reconcile transactions using fallback logic. • Optimised Django APIs to reduce latency by 60% and enhanced extraction accuracy with targeted prompt design. • Delivered real-time data sync pipelines transferring 500MB+ in under 30 seconds across services. • Experimenting with AI-powered SaaS solutions focused on reporting automation and enterprise workflow optimisation.

Location

Bengaluru

AI stack

anthropic apis

antigravity

claude code

claude

codex

cursor

Work experience

Software Developer Engineer - 2 at Karbon Business, Jul 2026 - Present · 1 mo, Bangalore

Software Developer Engineer - 1 at Karbon Business, Jul 2025 - Present · 1 yr 2 mos, Bengaluru South, Karnataka, India, • Designed and deployed fault-tolerant event processing framework with dynamic routing and automated recovery, enabling exactly-once consumption across AI microservices. • Built agentic workflows using OpenAI SDK + LangGraph with AI fallback paths, improving reconciliation reliability for high-volume financial transactions. • Optimised Django APIs with query restructuring, caching, and N+1 elimination leading to a 60% latency drop under production traffic. • Architected a real-time accounting sync system handling 500MB+ per push in <30s, integrating Flutter desktop clients with backend services. • Established CI/CD pipelines (Windows & backend) using GitHub Actions, automating build packaging, signing, and S3 distribution.

Full-stack Developer at Karbon Business, Jan 2025 - Jun 2025 · 6 mos, Bengaluru, Karnataka, India, • Prototyped agentic LLM pipelines and MCP-style experimentation frameworks to benchmark AI reliability vs latency in production-scale flows. • Built POC for document classification using Microsoft Graph APIs, enabling structured bill/bank statement identification for downstream services. • Developed evaluation frameworks to track accuracy and failure patterns of AI-based extraction and enrichment models. • Added unit/integration test coverage across services, uncovering edge-case failures and improving stability pre-deployment. • Contributed to microservice enhancements using FastAPI/Django with hands-on work in AWS infrastructure, Redis, and event-driven messaging.

Education

Master of Computer Applications - MCA, Computer Science, Christ University, Bangalore

Bachelor of Computer Application, Computer Science, Presidency College, Bangalore

Public links

github: https://github.com/vansh132

medium: https://medium.com/@shah.vansh132

vanshah-dev.vercel.app: https://vanshah-dev.vercel.app/