Headline

AI Engineer · RAG Pipelines · LLM-Powered Systems · Python Backend · LangChain · AutoGen

About

I build production-grade LLM and agentic AI systems the kind that run in real environments, not just notebooks. FEATURED PROJECT Built an end-to-end multi-agent LangGraph system for SRE incident response, deployed on Google Cloud Platform. A supervisor routes incidents to three parallel worker agents (logs, metrics, deploys), an aggregator synthesizes evidence, a critic loop demands revisions until satisfied, and a finalizer commits a structured remediation plan shipped a FastAPI service on Cloud Run with IAM-authenticated Cloud SQL state persistence, LangSmith observability, and Prometheus metrics. Evaluated across 12 incident fixtures with 7 evaluators (5 deterministic + 2 LLM-as-judge via Gemini 2.5 Pro) surfacing three measurable calibration biases documented in a public eval report. → github.com/harshithaneelam/sre-incident-copilot At Cognizant, Designed an LLM pipeline that identifies slow-running SQL stored procedures using AppDynamics REST APIs, retrieves source code via GitHub integration, and uses an LLM to rewrite and optimize query logic validated with a pre/post execution metrics framework in a production environment. OTHER PROJECTS → A multi-agent RAG chatbot using LangChain + AutoGen + Azure Cosmos DB (vector store) enabling semantic search and context-aware LLM responses over a product catalouge via a Flask API backend. → Real-time anomaly detection deep autoencoders on NSL-KDD, live Streamlit monitoring dashboard. My stack: Python · LangGraph · LangChain · AutoGen · RAG · Gemini (Vertex AI) · FastAPI · Cloud Run · Cloud SQL · LangSmith · Prometheus · Azure Cosmos DB · Flask · REST APIs · SQL · TensorFlow Always open to connecting with engineers and teams working on interesting LLM and agentic AI problems. harshineelam.30@gmail.com

Location

India

Work experience

AI Engineer (Programmer Analyst Trainee) at Cognizant, Aug 2025 - Present · 11 mos, Chennai, • Designed and implemented an end-to-end LLM pipeline to detect and optimize slow-running SQL stored procedures integrating AppDynamics REST APIs for performance data ingestion, GitHub API for source code retrieval, and an LLM to analyze and rewrite query logic. • Built a performance validation framework comparing pre- and post-optimization execution metrics, applying AI observability principles to measure real-world impact in a production environment. • Developed Python-based backend services to fetch, process, and store application performance metrics from AppDynamics into databases, focusing on reliability, efficient API usage, and debugging live systems. • Contributing to backend REST API development for multi-module enterprise workflows, with involvement in code reviews, modular architecture design, and clean code practices.

AI Intern at InTrain Tech, Oct 2023 - Nov 2023 · 2 mos, Built and trained a CNN for multi-class digit classification using Python and TensorFlow. Implemented end-to-end data preprocessing pipelines and applied hyperparameter tuning to improve model generalization accuracy.

Education

Bachelor's degree, Computer Science with Specialization in Artificial Intelligence and Machine Learning, VIT-AP University

Narayana Junior College - India