Tapas Thakkar — Software Engineer

I don’t just build AI demos; I architect production systems that scale.Currently, I lead AI initiatives at Jeeves, where I turned a manual document review process into an automated intelligence platform processing 50,000+ financial documents monthly with 95%+ accuracy.My focus isn't just on "making it work", it's on making it viable. I specialize in bridging the gap between cutting-edge AI research (like Anthropic’s Model Context Protocol) and secure, cost-effective enterprise software.WHAT I BRING TO THE TABLE:🚀 Production-Grade RAG: I’ve scaled document intelligence platforms from zero to 100K+ daily API calls across 5 countries, replacing 30-minute manual workflows with 30-second automated pipelines.💰 Cost Engineering: AI is expensive if you don't watch it. I build LLM Gateways and intelligent routing systems that dynamically switch between GPT-4, Claude, and Llama, slashing inference costs by 40% ($50K/mo savings) without sacrificing quality.🤖 Agentic Systems: I am an early adopter of Multi-Agent Systems and the Model Context Protocol (MCP), enabling AI agents to securely query SQL databases and perform complex financial analysis without hallucinating schemas.🛡️ Hard Engineering Foundation: Before the AI boom, I spent years building secure APIs (Google Apigee), optimizing search (Solr), and integrating complex Fintech ERPs (Xero/QuickBooks). I know how to lock down an API gateway just as well as I know how to prompt-engineer a model.CORE TECH STACK:AI/LLM: LangChain, LlamaIndex, AWS Bedrock, Vanna AI, Vector DBs (Qdrant/OpenSearch)Backend: Python (FastAPI), Node.js (NestJS), TypeScriptScale: AWS Lambda, Docker, Kubernetes, RabbitMQ, RedisI am always open to discussing High-Scale AI Architecture, Agentic Workflows, or Fintech Innovation.

Stackforce AI infers this person is a Fintech expert with a strong focus on AI architecture and backend development.

Location: Rajkot, Gujarat, India

Experience: 8 yrs 11 mos