Data Scientist · Bath, UK

Hi, I'm Mohsin.

I build machine learning and LLM tools that help public services make better decisions. Right now that means risk and forecasting work for adult social care, housing and children's services at Somerset Council.

Email me LinkedIn GitHub

About

I moved from India to England in 2023 for my Master's at Sheffield Hallam, did a year in industry at Wessex Water, and joined Somerset Council as a Data Scientist in July 2026.

Most of my work sits where machine learning meets messy, sensitive real-world data. I care about models people can actually trust: explainable outputs, clear data lineage, and pipelines that keep running after the demo.

Experience

Data Scientist, Somerset Council

July 2026 – present · Strategy, Performance and Communication

  • Leading a risk prioritisation tool for Adult Social Care: waiting list ranking, deterioration detection and demand forecasting.
  • Using Azure OpenAI inside a governed environment to pull structured indicators from free-text case notes, with a focus on prompt design and validation so results are consistent and explainable to caseworkers.
  • Data scientist on placement forecasting for Children's Services to support medium-term financial planning.
  • Building end-to-end ML pipelines with CI/CD, plus model governance and data lineage.
PythonSQLAzure ML Microsoft FabricAzure OpenAIMLOps

Data Specialist (industry placement), Wessex Water

One-year placement · part of MSc

  • Leakage detection using network flow and pressure data.
  • River water quality analysis.
  • Data pipelines on Azure.

Projects

Multimodal knowledge-graph RAG over scientific papers

Independent project · June 2026

A retrieval-augmented generation pipeline that reads scientific paper images, builds a knowledge graph from them, and answers questions with evidence. Built against a timed challenge deadline.

  • Tesseract OCR ingestion and spaCy entity extraction
  • Knowledge graph of about 11,800 nodes and 17,700 edges (NetworkX)
  • Text and image embeddings with MiniLM and CLIP

0.97 vs 0.53 mean evidence recall, graph-enhanced vs baseline, on a 10-question evaluation.

Small evaluation set, so treat it as a directional result rather than a benchmark.

View on GitHub →

Medical Recommendation System

MSc dissertation

A web app that takes reported symptoms and health data and suggests likely conditions and personalised recommendations. It mixes machine learning with rule-based logic, and I built it end to end, from model to interface to cloud deployment on Heroku.

View on GitHub →

Education & certifications

Education

MSc, Sheffield Hallam University

Includes a year in industry at Wessex Water

B.Tech Computer Science, Amity University, Noida

Microsoft certifications

  • DP-100 Azure Data Scientist
  • AI-102 Azure AI Engineer
  • DP-600 Fabric Analytics Engineer
  • DP-700 Fabric Data Engineer
  • PL-300 Power BI Data Analyst

Get in touch

Happy to chat about data science roles, research, or projects in public services and health.