Scroll

Available, Spring 2026Bloomington, IndianaAI & Data Engineer

UJJWALREDDY

03Next generation graph databases.

AI & Data Engineer. I build autonomous agents, RAG systems, and data pipelines with real production architecture — APIs, queues, observability, and developer experience that hold up. Not demos.

01About

I graduated with a degree in Data Science from Indiana University, with a focus on AI agents, data engineering, applied machine learning, and LLM-powered systems.

My work sits at the intersection of AI infrastructure, scalable ELT pipelines, and production-style ML tooling. I like building systems that are not just demos, but have real architecture behind them: APIs, databases, queues, observability, workflow orchestration, and clean developer experience.

AI Agents & Infra

MCP Servers · Orchestration · Trust Layers

Data Engineering

Airflow · dbt · Dagster · Kafka

Applied ML & LLMs

RAG · Embeddings · Evaluation · PyTorch

Developer Tooling

Local-first tools · CLI Security · Python · Go

02Experience

A short log of the work.

DEC 2025 → MAY 2026

Research Assistant

Indiana University Bloomington

UNDER PROF. DA YAN

01
Benchmarking TigerGraph, Neo4j, PuppyGraph & GraphFrames on the StackOverflow dataset, 763.7M edges, 8 vertex types, 10 edge types.
02
Authored 6 cross-system benchmark queries in GSQL, Cypher, and Spark SQL.
03
Discovered Neo4j loaded 88,692 fewer edges due to strict referential integrity, documenting a fundamental behavioral difference across graph database systems on incomplete real-world data.
04
Engineered ETL pipelines aggregating 10+ streaming + batch sources into a unified analytical data lake for researchers.

FEB 2022 → MAY 2022

Deep Learning Researcher

AIR Center, VIT-AP University

UNDER PROF. SIBI CHAKKARAVARTHY

01
Built Android malware detection system using Random Forest on permissions + activities (Androguard), 95% accuracy on 1,000 real-world samples. Published at ICISVL 2022. Springer.
02
Serverless real-time pipelines on AWS automating metadata feature engineering to drastically reduce fraud detection latency.
03
Python preprocessing microservices ensuring reproducible experimentation across 1,000+ APK datasets.

03Selected work

An archive of shipped things.

IdxProjectDisciplineStack

04Research

A short reading list.

Ten years of generative adversarial nets (GANs):a survey of the state-of-the-art

IOP MLST2024

Safeguarding Public Spaces: Unveiling Wallet Snatching through Edge Impulse Technology

IEEE RMKMATE2023

A Study on Android Malware Detection Using Machine Learning Algorithms

Springer ICISML2023

Certifications

Introduction to Model Context Protocol

Anthropic · Mar 2026

Developing LLM Applications with LangChain

DataCamp · May 2025

Create Machine Learning Models in Microsoft Azure

Microsoft · Jun 2023

Machine Learning

DeepLearning.AI · Jun 2023

Introduction to Computer Vision and Image Processing

IBM · Jul 2020

05Open source

0K★

GitHub stars on LangChain, where a merged PR now ships to OpenAI, Google, and every major AI company. Plus more PRs across LangChain core and dbt.

LangChain & dbt-core

PR #36462 merged into LangChain (133,000 GitHub stars, used by OpenAI, Google, every major AI company). Fixed a broken decoding example in filter_messages that caused copy-pasted official documentation to produce invalid code.

Additional PRs

3 additional PRs submitted to LangChain core and 1 open in dbt-core, all bugs found organically while building real projects, not from issue hunting.

06Agent report

How I build with agents.

A read on my working patterns across 43 sessions — pulled from my Paxel report. Hover a card for the full read.

Which kind of builder are you?GeneralistNo single working pattern dominates — you adapt your approach to the task.

How do you work with your agent?A back-and-forthYou work in dialogue, shaping the work as you go.

Your longest single session?16h 0mYour deepest uninterrupted stretch with an agent.

How often do you plan first?49% in plan modeYou opened 21 of 43 sessions in plan mode before writing code.

How many agents do you run?2 at onceAs many as 2 main sessions running at the same time.

How often do you change course?5% of the timeYou stop and redirect the agent mid-task rather than letting it run.

How polite are you?8 thank-yousYou thanked your agent in 8 of your messages.

How long are your prompts?240 words on averageMostly conversational prompts.

How much time did you put in?61 hoursAcross 43 sessions.

Your most cryptic prompt?"its reset start now"Somehow the agent knew exactly what you meant.

How much do you talk to your agent?25 prompts a session

07Beyond the code

Contributions · all time@ujjwalredd

—contributions · all time

On repeat

2002

Anne-Marie

Preview

Also in rotation
Finding Her
Kushagra, Bharath, Saaheal
Daylight
David Kushner
Heat Waves
Glass Animals

Daily driver

Pythonthe language I reach for

Off the clock

Cricket · F1 · Filmsand every other sport

Monsterand hard problems

UJJWALREDDY

I build systems that are not just demos. They have real architecture behind them.

AI Agents & Infra

Data Engineering

Applied ML & LLMs

Developer Tooling

A short log of the work.

Research Assistant

Deep Learning Researcher

An archive of shipped things.

RAG Pipeline with Hybrid Search

Meerkat

OmniCore-Local

Axiomeer

PromptPatch

Verix

Autonomous AI Company OS

LodeAI

Autonomous Supply Chain Control Tower

ELT Pipeline: Airflow + dbt + Postgres

Blackjack Analytics Pipeline

Structured Knowledge Notation

Algorithmic Trading System with AI Analysis

Route-Explanation-System

LLM-Fine-Tuning

IMDB Sentiment Analysis

UAutoml

Preflight