# Arsh Kumar

> Arsh Kumar is a machine learning engineer and software engineer (MLE + SWE), BEng Computer Science (First Class), University of Edinburgh, 2026. Works across the bigger AI stack: data, model, training, serving and the product on top. Experience in computer vision and diffusion models (Nethermind), financial time-series forecasting (with Prof. Jeremy Bejarano, UChicago), retrieval over research papers (ResearchRAG) and AI safety research (XLab, UChicago).

The full site content, as markdown, is at https://arshkumar.com/index.md. Contact: hi@arshkumar.com.

## Profile

- [Full profile (markdown)](https://arshkumar.com/index.md): selected work, projects, skills, CV and contact in one file
- [Website](https://arshkumar.com/): the same content with interactive 3D visuals

## Live projects

- [ResearchRAG](https://search.arshkumar.com): find research papers (arXiv and OpenAlex) by conceptual similarity to an idea; a generated "core idea" per paper retrieved 21% better than simple abstract embeddings
- [Games](https://play.arshkumar.com): pick two numbers, which act as coordinates that choose one of twenty Quick, Draw! classes, and Sketch-RNN, a recurrent network, draws a doodle stroke by stroke, in the browser
- [Contract Risk Scanner](https://contracts.arshkumar.com): upload a contract and see the risky clauses highlighted; Qwen3.5-2B fine-tuned with LoRA on CUAD, served on Modal behind a Cloudflare Worker

## Papers and code

- [Time-series forecasting methods for financial markets (OFR working paper, 2026)](https://www.financialresearch.gov/working-papers/2026/08/25/time-series-forecasting-methods-financial-markets/): forecasting benchmark built in collaboration with Prof. Jeremy Bejarano; Arsh wrote the initial OOP model APIs, GPU/workflow tests and data pre-processing for the repo.
- [ftsfr on GitHub](https://github.com/jmbejara/ftsfr): code for the forecasting benchmark

## Contact

- [Email](mailto:hi@arshkumar.com)
- [LinkedIn](https://www.linkedin.com/in/kumararsh/)
- [GitHub](https://github.com/thearshkumar)
- [X](https://x.com/__arshkumar__)

*AI-generated from my content and reviewed by me.*