Harsha vardhan Reddy

I’m a software engineer focused on building production-ready Generative AI systems, complex RAG pipelines, and multi-agent workflows. I turn experimental LLM prompts into predictable, stateful applications that stay reliable in production, not just in demos.

At Capgemini, I built an internal Multi-Agent System (MAS) called AgenticGallery using Python and LangGraph, enabling agents to pass state and automate complex corporate workflows while cutting manual overhead for the Innovation team. I also designed and deployed advanced RAG pipelines in Azure AI Foundry, using Azure AI Search with hybrid vector retrieval, semantic chunking, and custom system prompts to reduce hallucinations when reading messy, unstructured documents. I manage LLM deployments, token tracking, and evaluations to keep internal AI apps cost-effective and accurate.

My full-stack background means I don’t stop at the AI backend—I integrate it with scalable cloud infrastructure on Azure and deliver a clean, optimized user experience. I own AI monitoring dashboards end-to-end and push for strict code reviews and testing protocols, aiming for zero critical bugs reaching production; earlier, I strengthened performance and security by restructuring MongoDB schemas and indexes and built reusable React component libraries at Bhrish Labs.