I'm a concurrent BS/MS student in Computer Science & Data Science at Purdue ('27) who got tired of AI demos that never leave a Jupyter notebook. So everything I build has a destination: a children's hospital, a satellite, a DoD edge device, a Fortune 100 data platform, or a peer-reviewed venue.
My work lives at the intersection of deep learning and systems engineering — designing the model and the infrastructure that makes it survive the real world.
class Ojas:
education = "BS/MS CS + Data Science, Math minor @ Purdue ('27)"
focus = ["GenAI & LLM systems", "computer vision", "distributed ML infrastructure"]
industry = ["GenAI @ Amex 💳", "edge AI @ Northrop Grumman 🛡️"]
publishing = ["NeurIPS 2025", "IEEE (in progress)", "American Heart Association (in progress)"]
philosophy = "If it can't ship, it doesn't count."| Where | What's actually happening | |
|---|---|---|
| 💳 | American Express — GenAI SWE Intern | Architecting an LLM-powered Knowledge Graph across 10,000+ enterprise data tables, so AI agents can run natural-language queries without hallucinating schemas. Goodbye, hand-written YAML. |
| 🛡️ | Northrop Grumman — AI SWE Intern | Building defect-detection vision pipelines (ViTs + CNNs) for flexible hybrid electronics, compressed with TensorRT to run offline on rugged NVIDIA Jetsons in denied/degraded environments. DoD-funded NextFlex program. |
| 🫀 | Eli Lilly × Purdue BME — Research Assistant | Engineered AR-Net, an anatomy-aware spatiotemporal network that tracks the aortic root in pediatric echocardiograms — 97.9% expert-level correlation. Now deploying a clinical GUI to Yale New Haven & Riley Children's Hospitals to help catch fatal aneurysms early, with an American Heart Association paper in progress. |
| 🎵 | AI for Musicians @ Purdue — Research Lab Team Lead | Leading an NSF-funded team of 20 advancing automatic music transcription. Our evaluation standard was officially adopted by MIREX, and our work was published at NeurIPS 2025 — with an IEEE paper on the way. |
| 🛰️ | Astrobotics @ Purdue — ML Team Lead | Leading 14 engineers building autonomous robotic arms for on-orbit satellite servicing, with hardware testing partnerships at NASA and Vast. 7-DoF pose estimation, RL agents, 93% multi-step task precision. |
📜 Previously...
- Concrete Engine — Junior Software Engineer: HPC storage orchestration, Docker-isolated multi-tenant compute, and distributed multi-node training pipelines for large-scale AI workloads
- Qualcomm — Computer Vision Data Analyst Intern: optimized low-power vision models for Snapdragon X Elite edge devices; cut false positives by 39%
- Improved Google's MT3 music transcription model by ~15% on overlapping instruments with custom Viterbi-based decoding
- Built a 19-layer CNN that watches guitar videos and converts them into playable tablature (91% frame-wise accuracy)
| 📄 3 | 🫀 97.9% | 👥 34 | 🎵 1 | 🏆 2026 |
|---|---|---|---|---|
| papers — NeurIPS 2025 published, IEEE & AHA in progress | agreement with board-certified cardiologists | engineers led across two teams | industry eval standard, adopted by MIREX | CRA Outstanding Undergraduate Researcher HM — national honor |
