Hyderabad, Telangana

Former Product Analyst Role

Available for opportunities

Hemanth Chunduru

AI Systems Engineer

Designing LLM systems with architectural depth—retrieval pipelines, evaluation infrastructure, and latency-aware backends built to operate reliably at scale.

View Work ↓ Download Resume ↗ LinkedIn ↗

Experience

Prior Role

Product Analyst

Enterprise Software Environment

Gathered and formalized product requirements across engineering, operations, and business stakeholders—translating ambiguous inputs into structured specifications.
Identified process bottlenecks and redesigned workflows to reduce manual overhead, using data to validate before and after states.
Acted as a liaison between technical teams and business owners, maintaining alignment on scope, priorities, and delivery timelines.
Produced system documentation, decision logs, and analytical reports used to inform product direction and resource allocation.

Transition

Working inside product systems made the gaps obvious—decisions were made without reliable data pipelines, tooling was brittle, and no one had a clear model of how information moved through the organization. That led to a deliberate move toward the technical layer: learning how systems are actually built, how data is retrieved, and how models can be integrated without creating new fragility. The shift from product to engineering wasn't a pivot—it was a depth increase on the same underlying problem.

Selected Work

Company Knowledge Base RAG System

RAG · Semantic Search · Embeddings · Evaluation

GitHub ↗

Built an AI-powered chatbot that lets employees instantly query internal company documents — HR policies, onboarding guides, and SOPs — using natural language. PDFs are chunked, embedded into a FAISS vector store, and the top matching context is passed to GPT-4o, which generates grounded answers with source citations. Designed to fail honestly — if the answer isn't in the knowledge base, it says so rather than hallucinating. Evaluated using Ragas with a faithfulness score of 0.91.

0.91 Faithfulness Score

~2.3s Avg Response Time

Zero Hallucinations on Context

Python LangChain FAISS GPT-4o FastAPI RAG Ragas Evaluation

Social Sentiment Analyzer

NLP · Sentiment Analysis · Emotion Detection · LLM

GitHub ↗

Built an AI-powered dashboard that analyzes social media comments at scale. Upload a CSV export from Instagram, Twitter, or any platform — GPT-4o analyzes every comment for sentiment (positive, negative, neutral), emotion (happy, angry, sad, surprised), and a confidence score. Returns an AI-written executive summary, interactive charts, highlights for most liked and most polarizing comments, and a fully filterable per-comment table. Built for the enterprise workflow where social data is exported and analyzed in bulk.

4 Emotions Detected per Comment

GPT-4o Per-comment Analysis

CSV Enterprise Input Format

Python GPT-4o FastAPI Pandas Chart.js NLP Sentiment Analysis

Engineering Philosophy

Most AI work ships prototypes. I engineer the layer beneath— where retrieval fails silently,
latency compounds, and evaluation is the only truth.

A model is a component. The system is the retrieval architecture, the prompt contract, the fallback behavior, the evaluation loop. Those decisions determine whether something actually works—or just demos well.

Tech Stack

Python Backends, pipelines, data processing

LangChain RAG pipelines, chains, retrieval

Vector Databases FAISS · embeddings · similarity search

LLM APIs GPT-4o · prompt engineering · RAG

FastAPI REST APIs · async backends · routing

Pandas Data processing · CSV · NLP pipelines

RAG Systems Chunking · embeddings · retrieval · eval

HTML / CSS / JS Frontend UIs · dashboards · charts

Git / GitHub Version control · project showcase

Current Focus

Improving retrieval faithfulness in multi-hop RAG pipelines through structured query decomposition and context verification layers.
Building offline evaluation frameworks that measure answer groundedness and retrieval precision before changes reach production.
Exploring agentic system design patterns where tool selection, memory, and fallback behavior are treated as architectural decisions.

Ask Hemanth's AI

Have a question?
Ask directly.

This AI knows everything about Hemanth — his projects, transition story, tech stack, and work history.

Answers appear here

Contact

Get in touch

Open to AI engineering roles, collaborations, and interesting problems. Fill out the form and I'll get back to you.

→ chunduruhemanth2002@gmail.com