Skip to content

Business & Data Analyst — London, UK

Finding the signalin the noise.

I'm Ashutosh Parab — an MSc Business Analytics graduate who turns messy data into decisions, from crypto fraud detection at 96.2% ROC-AUC to macroeconomic forecasting.

Download CV

About — Click a concept

The page demonstrates what I do.

I hold an MSc in Business Analytics and Decision Sciences from the University of Leeds, built on a Computer Science engineering foundation. My work spans fraud detection, forecasting, and customer analytics — RFM segmentation, CLV modelling, behavioural analysis — always with the goal of an insight someone can act on.

Rather than list my skills, this section shows them. Each concept below re-arranges the same data points into the idea being named — because analysis is about making structure visible.

VARIANT AVARIANT BINGESTENGINEERTRAINDECIDE

Unlabelled points resolve into behavioural clusters — the basis of targeted strategy.

Selected Work

Projects

Dissertation research on wallet-level fraud in peer-to-peer cryptocurrency transactions using the Elliptic++ dataset. LightGBM classifier with SHAP interpretability achieved 96.2% ROC-AUC across 39 validated fraud indicators, and proved fraud is instantaneous (98.3% single time-step) rather than evolving over time.

96.2%

ROC-AUC

39

Fraud indicators

98.3%

Single time-step

PythonLightGBMSHAPElliptic++Statistical Testing
View on GitHub

Machine learning pipeline predicting Initial Coin Offering success. Compared Random Forest, XGBoost, and Logistic Regression with engineered features and hyperparameter tuning, evaluated on ROC-AUC to identify the strongest signals of fundraising success.

3

Models compared

ROC-AUC

Evaluation

PythonXGBoostRandom ForestScikit-learn
View on GitHub

ARIMA and Prophet forecasting models for monthly US Personal Consumption Expenditure, decomposing trend and seasonality and surfacing macroeconomic insight through interactive Power BI dashboards.

Monthly

Horizon

ARIMA + Prophet

Models

PythonARIMAProphetPower BITime Series
View on GitHub

K-Means and RFM-based clustering of e-commerce customers into behavioural segments, enabling targeted marketing strategy with clearly differentiated value, frequency, and recency profiles.

K-Means + RFM

Method

PythonK-MeansRFM AnalysisPandas
View on GitHub

Dissertation — University of Leeds · 2025 · Supervised by Dr. Liz Mason

Illicit funds don't hide in big moves.They hide in small, fast ones.

My MSc dissertation investigated wallet-level fraud in P2P cryptocurrency transactions on the Elliptic++ dataset — engineering 39 behavioural features across 1.27M records and applying SHAP explainability to 28,601 flagged wallets. A LightGBM classifier with time-aware train/test splits reached 96.2% ROC-AUC.

The key finding: high-frequency, low-value transactions are the primary mechanism for masking illicit fund flows, with transaction cadence and counterparty diversity the strongest predictive signals. A temporal ARIMA overlay turned this into early-warning fraud signals 2–4 weeks ahead at ~0.80 confidence — directly applicable to AML compliance and transaction monitoring in fintech and RegTech.

Read the research
Elliptic++LightGBMSHAPAML

96.2%

ROC-AUC

1.27M

Records

2–4 wks

Early warning

Top fraud indicators · mean |SHAP| (illustrative)

Transaction cadence

Counterparty diversity

Mean transaction value

Txn frequency bursts

Wallet age at first txn

In/out flow ratio

Experience & Education

Business Analyst (Freelance)

Chandgad Farm Fresh — India

Segmented a 200+ customer base into 3 behavioural groups through RFM clustering and CLV modelling, shaping a differentiated marketing strategy that contributed to ~15% revenue growth and a diversification plan across 3 new product lines.

2025

MSc Business Analytics & Decision Sciences

University of Leeds — Leeds, UK

Advanced Forecasting, Machine Learning, Data Visualisation, Quantitative Analysis, Evidence-Based Consultancy, Advanced Decision Making. Dissertation: fraud detection in P2P cryptocurrency transactions (96.2% ROC-AUC).

2024 — 2025

BEng Computer Science

Visvesvaraiah Technological University — India

Foundations in programming, algorithms, and databases that underpin my analytics engineering.

2019 — 2023

Contact

Let's find yoursignal together.

Actively seeking Business Analyst and Data Analyst roles in the UK — particularly in fintech, RegTech, and anywhere data shapes decisions. Graduate Route Visa holder — no sponsorship required, available immediately.