Hello, I'm
Ankit Singh
|
Data Scientist with a PhD in Physics. I build ML models, data pipelines, and deployed apps, with a background in statistical modeling and terabyte-scale scientific computing.
Hello, I'm
Data Scientist with a PhD in Physics. I build ML models, data pipelines, and deployed apps, with a background in statistical modeling and terabyte-scale scientific computing.
I’m a Research Fellow at the University of Nottingham with a PhD in Physics and 10+ years working with terabyte-scale simulation data, statistical modeling, and HPC. Recent projects include an end-to-end MLOps pipeline with drift monitoring, deep learning for galaxy morphology, a dual-model classifier for 104K UK road accidents, an LSTM congestion forecaster on TfL sensor data, RAG chatbots, and quantitative finance tools.
I’m an ISO/IEC 42001 certified practitioner (AI Management Systems) and AI+ Foundation certified, with additional applied data science credentials from WorldQuant University. 12 peer-reviewed publications, 100+ citations.
Dual-strategy ML system on 151K UK road accidents: emergency response model achieving 92.4% severe recall and traffic management model with 81% macro recall using SMOTE+Tomek and ADASYN.
RAG chatbot using FAISS/ChromaDB vector stores and sentence-transformers to answer UK immigration questions from GOV.UK data.
PyTorch LSTM forecasting motorway congestion from real Transport for London sensor data, reading a 4-hour window of 15-minute volume and speed readings to predict the next interval. Congestion is the minority class, rising from 5.5% of training intervals to 25.6% in the chronologically held-out period. Benchmarked against majority-class, rush-hour and speed-threshold baselines: 0.977 recall, 0.765 F1, PR-AUC 0.893 against a no-skill 0.256.