Diksha Singh

I’m a Data Scientist and independent researcher specializing in ML reliability, data quality engineering, and algorithmic trading systems. I found NXIO to turn market intelligence into production-ready signal generation and backtesting.

In NXIO, I architected an end-to-end market intelligence system for NSE, BSE, and MCX with real-time data pipelines, automated feature engineering, anomaly detection, and alerting. I also built quantitative trading strategies for price prediction and risk assessment while ensuring SEBI compliance with the April 2026 Algo-ID mandate.

Alongside building, I do ML reliability research focused on label noise, prediction calibration, and feature stability. I prototype with reproducible experiment designs, publish research documentation and technical papers, and use ablation studies to validate what actually works.

My recent work includes an LQNE label-quality engine, a Prediction Confidence Decay Tracker for post-deployment calibration monitoring, and a Feature Importance Instability Detector for diagnosing SHAP/permutation ranking drift. I bring an engineering mindset—FastAPI, PostgreSQL, Streamlit, and robust ML diagnostics—so models can be trusted in production.