Jyothiradhitya Somayajula

Software engineer with hands-on experience designing Retrieval-Augmented Generation (RAG) and agentic AI systems. Built an end-to-end RAG pipeline (PostgreSQL/pgvector, Sentence Transformers, locally hosted Qwen2.5 via Ollama) with a formal evaluation framework scoring 97.9% Hit Rate and 0.894 MRR across 240 ground-truth questions. Currently building an agentic system at Pharmcrux that pairs conversational RAG with in-product automation for navigation and form-filling.

Solid backend foundation in Python, FastAPI, and PostgreSQL, with production experience integrating the Google Gemini API for automated failure analysis.

Software Engineer / Data Analyst - Pharmcrux - Hyderabad, India

(2025-06)

• Building an agentic AI system that pairs a conversational RAG chatbot with in-product automation for navigation and form-filling, backed by a local LLM (Ollama) stack.
• Built an AI-assisted failure-analysis tool using the Google Gemini API that parses browser logs, DOM snapshots, and screenshots to auto-generate root-cause reports, reducing manual QA triage effort.
• Built embedding-based retrieval and LLM response-generation components for an internal enterprise RAG / document-intelligence application.
• Implemented Change Data Capture (CDC) pipelines with Debezium, Kafka Connect, and PostgreSQL, keeping data synchronized in real time across internal applications.
• Wrote modular Python/Selenium automation to execute application workflows, expanding scalable QA test coverage.

B.Tech - Computer Science and Engineering - Mahatma Gandhi Institute of Technology (2025)