Hello,
A Data Scientist.A Data Analyst.
About
Delivered data-driven insights for three companies by transforming raw, unstructured data into strategic assets. I leverage advanced prompt engineering and AI-augmented workflows to rapidly reduce time to insights. My workflow prioritises ethical integrity through a strict commitment to GDPR compliance and the FATES framework for responsible AI.
Featured · Recognition
Recognised in the Significant Contribution Awardcategory by the University of Sheffield Careers & Employability Service — for work delivered at TrackMyMachines Ltd, a project funded by the South Yorkshire Innovation Program.
Awarded at the Data & Design Hackathon 2026, University of Sheffield — for SwimMetrics, an Olympic-readiness dashboard built for Aquatics GB across 2M+ historical athlete records in a single 72-hour sprint.
Experience
TrackMyMachines Ltd · University of Sheffield
Sheffield, United Kingdom
Department of Computer Science & Engineering, PES University
Bengaluru, India
Third Axis Ltd
Bengaluru, India
Education
The University of Sheffield
Sheffield, United Kingdom · Sept 2025 - Aug 2026
Among the top 100 universities in the world — a Russell Group institution recognised for research output across data, computing, and biomedical sciences. Coursework spans statistical learning, optimisation, large-scale data engineering, and applied machine learning.
PES University · Bengaluru, India
Specialization in Aerospace Engineering
Tech Stack
Primary tools shown in accent. Adjacent tooling kept current — picked up as the problem requires.
Projects
Hackathons, capstones, research, and industry — each chosen for the leverage real data produced on a real outcome.
Predictive models on high-dimensional TCGA-BRCA RNA-seq data — log2 normalisation, variance filtering, and PCA to isolate differentially expressed features. Regularised linear, ensemble tree, and dropout-regularised neural network classifiers cross-validated for ROC-AUC, precision, and F1. Explainable-AI attribution validated against Spearman's ρ and Kendall's W for consensus across model classes.
Built a single scikit-learn pipeline shared across training and a FastAPI serving layer, eliminating data leakage and train/serve skew — enforced with a test asserting a row scored alone matches the same row scored in a batch. Compared four models by 5-fold cross-validation, tuned XGBoost with Optuna to 0.783 test ROC-AUC, and set a cost-based decision threshold that raised defaulter recall from 63% to 70% while lowering expected cost. Shipped as a containerized REST API with SHAP explanations, a sex/age fairness audit, 14 pytest tests, and a GitHub Actions CI pipeline.
View on GitHubTwo-part virtual analyst simulation for British Airways. Built a route-type lookup table from 10,000 scheduled flights to forecast lounge demand across three tiers at Heathrow Terminal 3, then trained a RandomForestClassifier on 49k+ booking sessions — validated with stratified 5-fold cross-validation (ROC-AUC 0.73) — to show that purchase-lead time, not customer origin, is the strongest driver of booking completion.
View on GitHubEnd-to-end quant desk simulation for Citi: priced a coffee futures contract and call option three independent ways (cost-of-carry, Black–Scholes, 10,000-path Monte Carlo) and cross-checked they agreed, then reused that pricing engine to structure a participation note and a digital option for two client risk profiles, and mapped market, credit, and operational risk mitigations for the book.
View on GitHubYou're looking at it. Drop your CV into a folder, paste one prompt into Claude Code, and the AI reads your CV and builds your entire site — same Tesla Design System, same animations, zero design decisions. 44 files, fully typed, builds clean. Open-sourced on GitHub so anyone can have a portfolio like mine.
View on GitHubBuilt SwimMetrics for Aquatics GB during the Data & Design Hackathon 2026 at the University of Sheffield — 2M+ historical athlete records processed in 72 hours via EDA and statistical imputation, then deployed as an interactive full-stack Streamlit dashboard translating raw data into actionable scouting intelligence.
Read the storyEnd-to-end RAG pipeline over 1,900 UK grocery products (Open Food Facts) using BAAI/bge-small-en-v1.5 embeddings and a ChromaDB vector store — fully local, zero API cost. Deployed as a Streamlit chat UI with NutriScore metadata filtering, cited responses, and real-time latency display. Evaluated with RAGAS across 25 hand-crafted Q&A pairs: context_recall 0.45, answer_relevancy 0.75.
View on GitHubEngineered a fully operational drone from custom carbon-fibre fabrication — 60 kg total thrust, 10 kg payload, FEA-validated for 10 m impact, with semi-autonomous flight via Mission Planner.
Contact