Hello, I'm
Nirek Sharma
ML Fairness @ Upstart
I research how to make machine learning systems fair and equitable.
I'm a research scientist at Upstart where I focus on Machine Learning Fairness. I work on detecting and mitigating disparate treatments and impacts of ML models on protected groups.
My career has been spent working on ML systems to solve problems that help people. But as these models have become more capable, I've become more passionate about addressing the unintended effects of ML models.
A Feature-Level Adjuster to Debias Proxy Features
A first of its kind technique to mitigate disparate treatmenet risk in ML models. Developed and applied at Upstart, abstract available below
Read the abstract →Algorithmic Debiasing via Post-Processing
A novel technique for algorithmically debiasing general ML models — developed at Upstart and applied to core underwriting systems. Published on arXiv, 2025.
Read the paper →- Algorithmic Debiasing via Post-Processing
- Same Data, Different Conclusions: Radical Dispersion in Empirical Results When Independent Analysts Operationalize and Test the Same Hypothesis
- Model for HIV/AIDS Incorporating Pre- and Post-Exposure Treatments & Reproduction Number Derivation
- Development of a New Inter-Institutional Partnership to Assess Health Literacy Disparities in the Context of Kidney Cancer and Smoking