• Designed and owned the backend architecture for a multi-tenant chat platform handling concurrent LLM workloads across providers (OpenAI, Anthropic, Groq), with a focus on extensibility and reliability.

• Built event-driven services using Go, Python, NATS, supporting low-latency streaming, asynchronous execution, and fault-tolerant workflows.

• Designed and implemented an extensible “Agents Store” by integrating LLMs with Model Context Protocol (MCP) servers, enabling tool-driven workflows across databases, documents, and SAP systems.

• Worked closely with the founder and customers to ship production features, demo the AI platform, gather feedback, and iterate on system design.

• Developed and deployed a custom Postgres MCP server to expose structured application data for conversational querying within the BlueRequests platform.

• Delivered specialized agents such as a SAP Service Generator that created service artifacts from user specs and supported downstream SAP import workflows.

• Designed and built real-time, asynchronous chat and comments infrastructure for the BlueRequests release management platform using Go and NATS.

• Developed the initial version of BlueFunda AI assistant chat client that later evolved into the production multi-LLM chat experience.

• Built a platform for the early-stage startup to automate VC investment research and due diligence workflows using agentic LLM workflows.

• Reduced manual analysis time from hours to under 10 minutes using LLM-based automation and Retrieval-Augmented Generation (RAG) pipelines.

• Engineered entity-extraction and deep-research agents using LangGraph with OpenAI and Perplexity APIs to reduce manual analyst loops.

• Built a RAG-driven Document Q&A engine to deliver citation-backed deal insights and significantly reduce analyst effort.

• Developed a Doc2Md conversion pipeline integrating multiple parsers (MarkItDown, Docling, LlamaParse, etc) and LLM providers, benchmarking accuracy / latency trade-offs to achieve ~3x faster parsing throughput for pitch decks and investment documents.

• Worked closely with founders to shape AI system architecture and delivery priorities.

• Designed and owned the State Cache Manager (SCM) in Go, a critical distributed component of the AVI Load Balancer used in high-availability enterprise environments.

• Improved state processing that reduced state convergence latency by 42% in 25K+ virtual-service environments through concurrency, caching, and data-structure optimizations.

• Refactored legacy SCM services and APIs from Python to Go, reducing memory footprint by 30% and improving long-term scalability and testability.

• Built and maintained new backend services in Go using gRPC and Redis, operating under strict SLA, reliability, and performance requirements.

• Refactored the Config Version service for Global Server Load Balancing (GSLB) using Go worker pools and Redis caching, processing ~5x larger volumes of concurrent updates to reduce version drift and improve reliability under load.

• Collaborated with cross-functional teams to deliver production features used by large enterprise customers.

• Worked on the IPAM / DNS subsystem of VMware’s AVI Load Balancer, contributing to core networking and infrastructure functionality.

• Assisted in the development of Custom DNS support, enabling customers to integrate external IPAM / DNS providers such as Infoblox.

• Developed an end-to-end test plan to validate system behavior across integration points and failure scenarios.

• Improved security posture by containerizing control scripts, addressing an identified vulnerability in the existing setup.