omkar pote

Data management · Machine learning

OmkarPote.

Hi! I'm a graduate student at Purdue University, working with the wonderful Dr. Romila Pradhan. I'm interested in data management and in making machine learning pipelines more robust, explainable, and fair. I've just defended my M.S. thesis.

Before Purdue, I studied at IIT Madras and spent a few years building software, from payment services to systems processing billions of records a day. Somewhere along the way, I got increasingly curious about the data going into these systems.

Explore my work GitHub
hello, world.
Pixel-art portrait of Omkar Pote
ohm-kaar poh-tay

A bit of research.
A bit of building things.

01 / Research & projects

A mixed assortment
of projects.

From machine learning research
to a potato and a snake.

Showing all 9 projects

Pixel-art balance scales holding two data cubes

Research / Purdue

Fairness through task-vector arithmetic

During my research assistantship at Purdue's Responsible Data Science Lab in 2024–2025, I worked with Dr. Romila Pradhan on task-vector arithmetic: using differences between model weights to try to reduce bias in machine learning models.

I presented this work at CERIAS and Purdue AARC Applied AI Day in 2025. Our CERIAS poster even won second place!

  • Pote, O. and Pradhan, R. “Unlearning Machine Learning Bias using Task Vector Arithmetic.” 26th CERIAS Annual Cybersecurity Symposium, Purdue University, 2025. Second Place, Poster Competition.
  • Pote, O. (Advisor: Dr. Romila Pradhan). “Resolving Machine Learning Bias using Task Vector Arithmetic.” Purdue AARC Applied AI Day, 2025.

I also spent some time exploring large model management and model lakes: how we keep track of models, where they came from, their different versions, and how to find the right one. That was exploratory work, separate from my thesis.

Pixel-art oscilloscope with a rewind arrow

Research / IIT Madras

Counterfactual explanations for time series

Before Purdue, I worked with Dr. Rajagopalan Srinivasan in the DART / iACE Labs at IIT Madras as a Project Scientist (October 2023–June 2024).

I worked on counterfactual explanations for time series: figuring out what would need to change in a signal for a model to make a different prediction. We used this to understand process faults and explore how to recover from them. This work led to a short paper at ESCAPE 2025.

Peer-reviewed short paper: Pote, O., Majji, D., Bhakte, A., Srinivasan, B., and Srinivasan, R. “Timeseries-based Counterfactual Explanations for Process Fault Diagnosis and Recovery.” 35th European Symposium on Computer Aided Process Engineering (ESCAPE), 2025. Book of Short Papers, pp. 235–236.

Pixel-art open book with code brackets

Teaching / Purdue

Teaching & academic service

In Fall 2025, I was a Graduate Teaching Assistant for CNIT 18000: Introduction to Systems Development at Purdue. I ran two lab sections on my own, handling the teaching, grading, and student support. We worked through systems-development and database concepts, along with the assignments that put them into practice.

I've also been an external reviewer for the IEEE International Conference on Data Engineering (ICDE 2025).

Pixel-art golden potato

Industry / Capillary

POTATO

During my three years at Capillary, my biggest project was POTATO (Over-The-Air Transactions Orchestrator), an in-house system for cashless fuel payments. It connected payment providers, fuel stations, and other services to make the whole transaction work.

The system supported more than 10,000 transactions per minute. Alongside the payment service, I worked on automating builds and deployments, and keeping an eye on how everything was running.
Why did I name it so?

Pixel-art database with a data pipeline

Industry / Capillary

ReOn

Before POTATO, I worked as a Big Data Engineer on ReOn, short for Relational Ontologies. It handled the data behind Capillary's customer analytics and business intelligence tools, processing more than 6 billion new records every day.

I worked on the Spark pipelines that moved and transformed this data, and the APIs around them. As an intern, I also built tools to run and monitor these jobs.

Pixel-art Count Dooku as a cuckoo bird, with a cloak and gold chain

Side project / Open source

Cuckoo Dooku

Everyone loves open source software. Everyone loves Star Wars (right?). Cuckoo Dooku brings the two together: a Star Wars-themed Chrome extension to help you stay productive.

It even got a commendation at the FOSS Hackathon in 2021. (Must be really cool.)

Pixel-art snake chasing an apple

Side project / C++

Snake (C++)

I started coding at eight with Logo and QBasic. In ninth grade, I picked what seemed like a simple one-day school project: building a snake game. (It's just like a Logo turtle, right? Wrong.)

It took weeks. But I learned a lot, and while plenty of my early projects have disappeared over the years, I've carefully preserved this one.

Watch on YouTube (opens in a new tab)

Pixel-art produce crate with a small gear

Industry / Earlier work

Earlier industry projects

At Tesco in 2019, I helped automate part of the food-donation process, working on tools to classify products and move that information through the system. The wider process served more than 2,500 stores and 6,000 charities.

A year earlier, I helped Crest IT Consulting with a new website and visual identity. A bit of coding, a bit of graphic design.

02 / The toolkit

Things I work with.

I like trying out new tools, but I usually end up spending just as much time figuring out how they work and why they work well for some problems but not others. Here's some of what I've worked with.

  • Primary Interests

    Data Management for ML, Data Quality, Preprocessing Pipeline Selection, Distribution Shift, Explainable AI, Algorithmic Fairness

  • Research Methods

    Metafeature Engineering, Clustering, Cross-Validation, Tabular and Time-Series Classification, Counterfactual Explanations, Fairness Evaluation, Task-Vector Arithmetic

  • Software Development

    System Design, Microservices, Spring Boot, REST APIs, Object-Oriented Programming, Testing, Git, CI/CD

  • Data Engineering

    ETL Pipelines, Apache Spark, AWS EMR, MySQL, MongoDB, Neo4j

  • Languages

    Python, SQL, Java, Scala, C/C++, Bash, MATLAB

  • ML & Data Tools

    NumPy, pandas, scikit-learn, PyTorch, Hugging Face Hub API, Sentence Transformers

  • Web Development & Design

    React, HTML/CSS, JavaScript, Flask, Django, Graphic Design

  • Infrastructure & Tools

    Docker, Kubernetes, AWS, RabbitMQ, Redis, Apache NiFi, LaTeX

03 / Beyond the keyboard

Let's talk?

I'm always up for a conversation, especially if it turns philosophical. I'd probably double-major in philosophy if I could go back in time. I like meeting new people, trying new food, and learning about different languages and cultures.

On weekends, I try to make time for soccer, though these days it's mostly watching. I also geek out on music, usually indie and alternative rock, with plenty of shoegaze and post-punk.

And if you're looking for low-effort humor, or an unnecessarily long discussion about how puns emerge from the abstraction of language, you've come to the right place.

Get in touch