CS @ University of Cincinnati · Building at Oumi
Aniruddhan
Ramesh
I like figuring out what's happening inside neural nets and building tools that make it easier to find out.
About
Figuring out what's inside the black box
I'm a CS student at the University of Cincinnati (3.99 GPA), originally from Abu Dhabi. Right now I'm at Oumi in Seattle, working on data synthesis and interpretability tools for foundation models.
My research on polysemanticity in language models was accepted at the NeurIPS 2025 Mech Interp Workshop. I also run Bearcat Ventures, UC's $1M student-led venture fund, and co-founded PhizzIO, a CV platform for physical therapy that won a few pitch competitions.
Mechanistic Interpretability
I want to know what's actually going on inside these models — which circuits fire, what features form, and why a model does what it does.
Open-Source AI Tooling
Most of my work at Oumi has been making it easier to train, evaluate, and ship foundation models. I think good infra is what makes good research possible.
Applied ML
I like taking ML into messy real-world domains — physical therapy, clinical notes, music transcription. The interesting problems live at the edges.
Experience
Where I've been building
Machine Learning Engineer Intern
OumiBuilding data synthesis and interpretability tools on Oumi's open-source foundation model stack. Currently putting together an MCP server integration from scratch.
AI Research Intern
AlgoverseStudied how polysemanticity emerges and stabilizes in language models. The work got accepted into the NeurIPS 2025 Mechanistic Interpretability Workshop.
Biomedical Informatics Software Engineer
UC College of MedicineWorking with the iCDCU lab building clinical software. Designed a research project management system for a pediatric heart institute. Two publications came out of this.
Managing Partner
Bearcat VenturesRunning UC's $1M student venture fund. Sourcing deals, leading diligence, managing a team of analysts. Took 2nd at the VCIC Midwest Regional at Carnegie Mellon.
Machine Learning Intern
Kinetic VisionBuilt ML models to analyze surgical staple structural integrity, plus the data pipelines feeding them.
Software Engineer Intern
Phillips Edison & CompanyTwo co-op rotations. Built an ETL pipeline using multimodal AI to pull structured data out of messy real estate documents like leases and site plans.
Publications
Research I've been part of
Recognition
A few highlights
Projects
Things I've shipped
PhizzIO
CV-powered physical therapy app — watches you do exercises and tells you if your form is off. Took 2nd at RevolutionUC.
Oumi Platform
My open-source contributions to Oumi — inference engines, eval tooling, synthetic data, and more.
ctrlFind
Search your codebase with plain English. Built in Rust — embeds code into vectors and finds what you're looking for with cosine similarity.
GraphRag
RAG but with a knowledge graph instead of flat chunks. Built for searching research papers where context and connections between ideas matter.
Mridangam Transcription
Listens to mridangam (South Indian drum) recordings and transcribes the stroke patterns. A niche problem I cared about solving.
Clinical Notes
Takes long, messy clinical notes and pulls out what actually matters for the patient and their doctor.
Contact
Let's talk
Happy to chat about ML research, open-source, startups, or whatever you're working on.
rameshad@mail.uc.edu