Basavraj Chinagundi
Hey, I'm an AI Software Engineer with a passion for building intelligent systems and scalable solutions. Currently pursuing my Master's in Computer Science at Arizona State University, I specialize in AI/ML, full-stack development, and data engineering.
I've built RAG-based chatbots, automated code generation pipelines, and agentic AI workflows. I love working with cutting-edge technologies like LangChain, PyTorch, and modern web frameworks to solve real-world problems.
Download ResumeEducation
Master of Science, Computer Science (Aug 2025 - July 2027)
Arizona State University, Tempe, AZ
GPA: 4.11/4
Technical Skills
Experience
AI Software Engineer | Deloitte
Oct 2023 – July 2025
- Built a chatbot that actually understands banking documents by figuring out how to chunk them based on token limits and process them asynchronously. Also automated the conversion of old Informatica XML mappings into SQL - basically turned weeks of manual migration into a button click.
- Connected four different AI tools to work together using LangGraph, so they automatically hand off tasks to each other. Also built ETL pipelines that pull data from GCP and Azure, then load it into Snowflake and Databricks without breaking.
Software Development Engineer Intern | Samsung SDS
Mar 2023 – June 2023
- Replaced manual inventory tracking with a Java/React microservice that handles 1,000+ products. Cut errors by 90% and saved the team 10 hours a week. Then used that data to build ARIMA and Prophet models that predict demand 30% better than before.
Featured Projects
STaR: Self-Taught Reasoner
View on GitHub (opens in a new tab) | Python, PyTorch, Hugging Face
Took the STaR paper and implemented it from scratch on Llama 3.2-3B. Started strong at 73.8% on math problems, but then watched it collapse to 64.9%. Spent time figuring out why - turns out the model was generating increasingly bad training examples for itself. The real project was debugging why self-improvement broke down.
Slapp-AI
View on GitHub (opens in a new tab) | Python, PyTorch, Multimodal Models, Supermemory
Tinder for clothes. You swipe, it learns. Used a vision model to understand what's in the images, then built a graph of your preferences that updates with every swipe. Indexed 5,800+ items so recommendations actually make sense.
ShadowSearch
View on GitHub (opens in a new tab) | JavaScript, Cloudflare Workers, Llama 3.1
Chrome extension that answers questions about whatever page you're reading. Runs on Cloudflare's edge so it's fast. The tricky part was keeping conversations separate for each tab while sharing the same context. Won a hackathon competing against 600+ people.
Connect
Email: basavrajchinagundi10@gmail.com
LinkedIn: linkedin.com/in/basavrajchinagundi (opens in a new tab)