Y. Siva Sai Krishna

Designed and deployed a multimodal AI pipeline for an AI-powered Security Design Review platform, ingesting PRDs and C4 diagrams and converting visual architectures into structured representations for downstream security analysis. Architected a continuously synced code layer to provide live, queryable code context for LLM reasoning. Built a context-fetching layer on the Claude Agent SDK to autonomously pull relevant source code, improving grounding and reducing hallucinations. Encoded 100+ recurring security and quality patterns from historical reviews to drive prompt iteration and reliability, and integrated 4+ third-party inputs into the knowledge pipeline.

Owned the data plane and LLM application layer for an AI-powered marketing intelligence platform ingesting 10+ sources across social, news, forums, and reviews. Architected and deployed a multi-agent AI system using CrewAI with role-specialized agents to transform raw signals into structured, decision-ready narratives. Designed distributed ingestion pipelines with FastAPI, RabbitMQ, Celery, and Airflow. Implemented LLM observability via Langfuse and built a grounded RAG pipeline using LangChain, Pinecone, and OpenAI/Google models to provide grounded, evidence-backed responses.

Led a team of interns to build an AI career guidance product with features like career matching and salary comparison, used by 500+ students. Helped design a Crew Marketplace enabling monetization for 200+ creators, including a creator admin dashboard, data collection scrapers, and a Chrome extension. Built multi-tenant systems and collaborated on a brand collaboration tool for fashion/beauty brands. Implemented data pipelines using Python, React, FastAPI, GraphQL, Node.js, RabbitMQ, Celery, Postgres, Firebase, and AWS; developed a secure file upload/auth system and a Chrome extension for bookmarks.