You're herefor a reason.I'm Sepehr Zarnegar.

Explore projects View CV
Sepehr Zarnegar in professional attire
Based inLondon, UK

01 / About

About me

Warwick Business School graduate (MSc in Information Systems & Digital Innovation) with proven experience across data analysis, business analysis, and digital transformation. I specialize in aligning business strategy with technology to translate complex data into high-impact, actionable outcomes.

02 / My story

My journey

Education and experience, brought together as one continuous story.

The next chapter

Future

What’s next?

2024 – 2025

Education

Warwick Business School

MSc Management of Information Systems & Digital Innovation

Warwick Merit Scholarship
May – July 2025

Experience

Allnex

Data & AI Strategy Consultant

Improved platform engagement by 60% and developed an internal AI assistant supporting 1,000+ employees.
August – October 2023

Experience

BAT Pars

Business Strategy Intern

Analysed 10+ brands and behavioural research from 200+ respondents to inform portfolio decisions.
August 2023

Education

University of Tehran

BSc Management

First Class Honours · Top 10% of cohort
May – July 2023

Experience

Eastcool

Business Technology Intern

Automated 500+ transaction rows per run and reduced a two-hour month-end process to under 30 minutes.

I didn’t come this far to only come this far.

03 / Projects

Selected projects

GitHub ↗

Crypto & equities research / 01

After the Bell

Can Bitcoin's weekend movement help explain the next opening gap in crypto-linked stocks? Explore a historical replay, make your own prediction, then compare it with an expanding-window forecast and the actual opening gap.

156 distinct weekends studied3 crypto-linked equities
PythonSQLTime-aware forecastingInteractive data storytelling
Try the interactive replay ↗GitHub repository ↗
Mean absolute opening-gap error by stock: Bitcoin-linked model versus an expanding historical-mean forecast. COIN: 1.314 versus 2.540 percentage points; MSTR: 1.288 versus 3.085; MARA: 1.377 versus 3.046. Lower is better.

Historical research, not a trading strategy.

Machine-learning audit / 02

E-commerce Purchase Audit

A responsible modelling audit that tests whether customer interactions can predict purchase and determines that better data is more valuable than deploying an unreliable model.

2,871 usable interactions0.468 test ROC–AUC
PythonMachine learningLeakage controlModel evaluation
GitHub repository ↗
Comparison showing the selected purchase model does not improve on the baseline

Evidence before deployment.

Business case study / 03

Cyclistic Bike-Share

An end-to-end analysis of 5.7 million rides, translating rider behaviour into focused recommendations for membership conversion.

5.7M cleaned rides56% longer casual rides
DuckDB SQLPythonData cleaningBusiness recommendations
View case study ↗GitHub ↗
Chart comparing average and median ride duration for members and casual riders

Behaviour translated into action.

04 / Contact

Let'sconnect.