Skip to content
Aadit Golwala

Aadit Golwala

ML and software engineer

SWE intern at Unlearn · ML research at Purdue ECE

I'm a computer engineering student at Purdue. Right now I'm at Unlearn, working on the models and data tooling behind clinical-trial digital twins, and in Purdue's ECE department building multimodal computer vision for a holographic display.

24agolwala@gmail.comGitHub ↗LinkedIn ↗

Work

  1. Unlearn

    2026 —

    Software Engineering Intern

    ML for clinical-trial digital twins: model architecture and tracking, data protocols and curation, agentic workflows, internal tooling.

  2. Perception Innovations · Purdue ECE

    2025 —

    Machine Learning Engineer

    Multimodal computer vision for the Lightform 1 holographic display — head and gaze tracking, gesture recognition from live video.

  3. Oro Labs

    2025

    Software Engineering Intern

    CI/CD and deployment pipelines in GitHub Actions, plus internal tooling. ~10% dev time saved, ~80% of routine fixes automated.

  4. DataMermaid

    2023

    Software Engineering Intern

    R and Python ETL pulling coral-reef survey data from several research platforms into one analysis surface through their APIs.

Selected projects

Multimodal music emotion recognition — case study

In progress2026

Reading the mood of a song from its score, its lyrics, and its audio — then measuring which of the three actually knows.

The hard partAudio hears energy but not mood, so the two quadrants where valence and arousal disagree collapsed to near-zero F1. Predicting valence and arousal as separate regressions and reading the quadrant off them — plus SMOTE on the minority classes — took macro-F1 from 0.40 to 0.49 and stopped the model ignoring the hard classes entirely.

  • 0.764 macro-F1 · GPT-5.5 zero-shot
  • 0.743 · audio + lyrics fusion
  • 5 approaches compared
Python / scikit-learn / PyTorch / DistilBERT / OpenAI APICode ↗Read →

Writing →

About

Most of what I build outside of research gets shipped rather than left in a notebook — an iOS app on TestFlight, an ACL-safe RAG service on AWS, a pair of smart glasses that read a homework problem out loud and answer it.

I came up through FRC robotics, which is why a CNC router and a 3D printer feel about as normal to me as a terminal. That is also how a posture-corrective backpack ended up going from a spec sheet to pre-production samples.