Software engineer, AI/ML · UT Austin ’27

Looking for new-grad roles in AI/ML and software engineering, winter 2026 onward.

I'm a computer science student at UT Austin, co-founder & lead engineer at DUIBL, where I shipped Stilez, an AI outfit-rating app, and previously a software engineering intern at Cox Automotive. I build AI systems, then prove they work; the work below comes with its numbers.

  • Applied ML / LLM engineeringRAG and retrieval, evaluation harnesses, fine-tuning and distillation
  • ML infrastructure & MLOpstraining and serving pipelines, drift monitoring, CI/CD for models
  • Software engineeringbackend and full-stack, mobile (React Native), production systems
See the work

Experience

Where I've shipped, newest first. Each stop opens a page with the details, the stack, and links.

  1. — now

    Co-founder & lead engineer Stilez

    DUIBL

    Co-founded DUIBL and built Stilez, an iOS app that rates your outfit with AI, owning the product from architecture through App Store launch.

    • −67%image-gen spend
    • 2×rating throughput
    • 68%cache hit rate

    React Native · TypeScript · Postgres · pgvector · Deno · LLM pipelines

  2. —

    Software engineer intern vAuto

    Cox Automotive

    Summer internship on vAuto systems used by 7,500+ dealerships: vehicle sourcing, deployment tooling, and an internal auditing app.

    • −90%load time
    • −80%blue/green shift time
    • 1,000+inconsistencies surfaced

    React · AWS · Slack API · Blue/green deployments · REST APIs

  3. 2024

    Earlier

    A summer in UT Dallas's VR lab building a wind-tunnel simulation in Unity, and co-founding Growth Prep Academy, an AI test-prep platform that reached 500 beta users.

Projects

Independent work on LLM reliability, MLOps, and distillation, each measured against a baseline.

  1. —

    LLM reliability harness

    Aegis

    A reliability harness that wraps any RAG pipeline so its answers are verified, cited, or withheld, and makes a small open model compete with frontier ones.

    • 19×lower operating cost
    • <10%injection success, from 80%
    • 40kgenerations, every claim cited

    Python · LangGraph · Qdrant · MLflow

  2. —

    Self-healing MLOps platform

    Demand Forecasting Platform

    An MLOps platform that forecasts retail demand, watches its own drift, and retrains and redeploys itself without anyone stepping in.

    • −46%forecast error (WMAPE)
    • +71%accuracy after a demand shock
    • 206automated tests

    Python · LightGBM · MLflow · DVC · FastAPI · Docker

  3. —

    Research

    Distillation Reasoning Transfer

    Research into whether a small local model can learn to reason from a frontier model using only its outputs, trained on consumer hardware.

    • 30→46%GSM8K accuracy
    • ~46%of the gap to the teacher closed
    • 3,000filtered reasoning traces

    Python · PyTorch · Hugging Face · QLoRA · PEFT

Toolkit

Languages
Python, Java, C++, C, C#, Go, JavaScript, TypeScript, SQL, HTML/CSS
AI / ML
PyTorch, TensorFlow, Scikit-learn, Hugging Face Transformers, OpenCV, NumPy, Pandas
Agents & LLMs
LangChain, LangGraph, Google ADK, RAG pipelines, multi-agent systems, vector databases (pgvector, Chroma, Qdrant), fine-tuning, distillation, evaluation harnesses
Infra & DevOps
AWS (S3, Lambda, RDS, ECS), Docker, Git, CI/CD, GitHub Actions, REST/Express APIs, React/Next.js, PostgreSQL
Education
The University of Texas at Austin — B.S. Computer Science, expected December 2027.
Data Structures & Algorithms · Operating Systems · Computer Architecture · Machine Learning · Artificial Intelligence · Natural Language Processing · Computer Vision · Generative Visual Computing

Say hello.

Open to conversations about roles, research, and anything involving LLM systems that need to be trusted in production. My inbox is open.

varunvsaran@utexas.edu