Headline

Agentic AI Engineer | Production LLM Systems & Multi-Agent Workflows | LangGraph | AI Automation

About

I am an AI Engineer specializing in agentic AI systems, multi-agent workflows, and practical automation solutions that solve real business problems. In my current role at Karbon Business, I work on building production-grade AI systems across document processing, checklist generation, extraction, enrichment, and intelligent decisioning. My experience includes designing asynchronous LLM workflows, tool-calling systems, state management, validation pipelines, and fallback mechanisms to make AI systems reliable, scalable, and production-ready. I have worked on products such as FlowTAX and AI Accountant, where I helped build end-to-end pipelines for structured data extraction, historical data enrichment from Tally, and AI-based accounting automation. I focus strongly on deterministic-first design, hybrid LLM + rule-based approaches, and robust orchestration to improve accuracy, speed, and consistency. I enjoy building systems that combine strong engineering fundamentals with modern AI capabilities to deliver measurable impact in automation, efficiency, and user experience.

Location

Greater Bengaluru Area

AI stack

langchain

langfuse

langgraph

mcp

Work experience

SDE 1 at Karbon Business, Aug 2025 - Present · 11 mos, Bengaluru, In my current role as an AI Engineer, I build and deploy agentic AI systems for automation, including multi-agent workflows for document processing and accounting use cases. I design end-to-end pipelines covering extraction, enrichment using historical data (Tally), and structured output generation with deterministic and LLM-based approaches. I also focus on scalability and reliability by implementing async orchestration, validation, and fallback mechanisms in production systems.

Generative AI Engineer at Karbon Business, Jun 2024 - Aug 2025 · 1 yr 3 mos, FlowTAX — Multi-Agent System for CPA Workflow Automation (US Tax Filing) •Designed and implemented a 5-agent system (checklist generator, doc classifier, data extractor, questionnaire generator, AI assistant), taking ownership of checklist, classification, and extraction agents end-to-end. •Reduced checklist generation from ~3 days (manual CPA effort) to <30 sec, while enabling document classification (~5 sec) and bulk extraction + calculation (~100 sec) using async parallel execution. •Built event-driven architecture using webhooks + Inngest for queueing, retries, long-running workflows, and stateful multi-step execution with tool-calling. •Implemented validation, recovery, and failure-handling mechanisms to ensure consistency across agent workflows.

Education

B.tech, CSE, Centurion University of Technology and Management

class xii, Science, NIST Higher Secondary School