00 / Portfolio — 2026Based in the Isle of Man

DominicHubble.

Applied AI  /  Software Engineer
BSc CS — Loughborough — class of ’26

Building data-driven systems where machine learning meets the messy edges of real products — from Flutter apps shipping to users, to NLP pipelines unpacking financial sentiment.

View work →Download CV
Portrait of Dominic Hubble
Open to roles
Dominic Hubble2026
Currently — Software Developer @ MezzeDataAwarded — First-Class Honours / 77.1%
github /dominichubbleemail dominichubble@gmail.com

01 / About
On the record.

Origin Walsall, United Kingdom
Based Isle of Man

Applied AI & Software Engineer with a working obsession for the seam where data, models, and product collide. I’ve shipped full-stack platforms, mobile apps in Flutter, and NLP pipelines — and led a six-person team through a complete delivery cycle.

Currently a Software Developer at MezzeData (part-time, remote), having just graduated from Loughborough with a First-Class BSc in Computer Science. Previously interned at PDMS across front-end, testing and UX, and ran Python automation pipelines for Loads of Traffic from 2022–2024.

My final-year work on financial sentiment analysis became a first-author journal paper, currently under submission. Outside the editor: hockey for the Isle of Man Men’s Squad and Loughborough Town 1s, reading on financial systems, and thinking about how AI changes real operational work.

4 years of shipping production code
6 engineers I’ve led on delivery
△ On the pitch

Hockey keeps me honest off the keyboard — competitive, team-first, and a useful reminder that the best systems are the ones that hold up under pressure.

Dominic Hubble playing hockey on a blue pitch in Loughborough kit
Match dayLoughborough
Dominic Hubble receiving an Isle of Man hockey Young Player of the Year award
Young Player of the YearIsle of Man

02 / Featured work
Selected projects.

8 projects
+ more on github
P / 01
Professional• In ProgressMezzeData, 2025–

MezzeDocuments
content system.

An in-house transformation layer that turns static PDFs into structured, interactive in-app content. Modular layers, data pipelines, and caching designed for delivery efficiency at scale — built end-to-end in Flutter/Dart with an architecture I own.

FlutterDartData pipelinesCaching layerArchitecture
SRCdocument.pdfRAW
↓
PARSElayered_segments[]CHUNKED
↓
CACHEedge_storeWARM
↓
OUTinteractive.viewSHIPPED
P / 02
Final-year project• Journal paper2026

Financial sentiment
analysis.

An explainability-guided study using FinBERT on multi-source investor text (Reddit, X, news). LIME traced a neutral-class failure mode; a targeted fine-tune recovered neutral recall by 18.4 points and lifted accuracy to 86.5%. Now a first-author paper under submission.

Macro F1 / accuracy   0.803 · 86.5%
PythonFinBERTBERTNLPLIMEPyTorchscikit-learn
P / 03
Academic• Team Lead2024

Productivity
& knowledge
system.

Led a 6-person team to deliver a full-stack system using React, MySQL, and GCP. Agile / Jira, shipped ahead of schedule, with task management and data-driven dashboards as first-class citizens.

Read the project write-up →
DHAHCKGMJTSRFM
ReactMySQLGCPJiraAgile
AI / Data Science

Machine Learning Model for Time-Series Prediction

Neural network models for time-series prediction with an end-to-end ML pipeline.

PythonNeural networksMachine learning
04
AI / Data Science

Data Mining

ML analysis pipeline on a ~5,000-row socioeconomic dataset covering cleaning, modelling, and interpretability.

PythonpandasSHAPscikit-learn
05
Product / UX

UniLink HCI App

User-centred university app concept built using HCI methods, survey feedback, and privacy-aware design.

HCIUX researchFigma
06
Product / Full-Stack

Student Enterprise Dashboard

Planned a student enterprise management platform with phased agile delivery.

ReactSupabaseAgile
07
Enterprise / Business Process

ERP Systems

SAP S/4HANA coursework covering procurement, production planning, sales, billing, and payment workflows.

SAP S/4HANA
08
More

Everything else on GitHub

Source, experiments, and work in progress.

github / dominichubble →

03 / Research
In print.

1 paper
first author
Submitted2026

Diagnosing and mitigating the neutral-class deficit in financial sentiment analysis: An explainability-guided approach on multi-source investor text

Dominic Hubble, Stephen Lynch · First author
Data Science in Finance and Economics (DSFE), AIMS Press — Special issue: Machine Learning in Economics and Finance

First-author journal paper that began as my undergraduate dissertation and is now being submitted to publishers: an explainability-guided method that diagnoses FinBERT's neutral-class weakness with LIME and repairs it via targeted fine-tuning, recovering neutral recall by 18.4 points (to 71.7%) and lifting accuracy to 86.5%, validated on a multi-source benchmark with paired significance testing.

Peer-reviewed journal article

04 / Experience
Track record.

3 roles
2022 — present
Three
positions
/ four years
/ shipping
real things.
Apr 2025 — Present

MezzeData Limited

Software Developer (Part-time)

Leading development of a Flutter-based mobile application, owning architecture, feature delivery, and technical direction.

  • Contributed to a production Flutter application, implementing mobile features across UI, navigation, caching, and CMS-driven content delivery.
  • Helped replace a native iOS application with a modular Flutter codebase, improving maintainability and cross-platform delivery.
  • Developed an in-house system transforming static PDFs into structured, interactive in-app content to improve usability and content delivery.
Aug–Sep 2024

PDMS

Software Internship / Work Experience

Cross-functional software internship spanning front-end development, business analysis, testing, UX, project management, and marketing.

  • Completed a cross-functional software internship, gaining experience across front-end development, business analysis, testing, UX, project management, and marketing.
  • Built exposure to React, Umbraco, and Hotjar within a commercial software environment.
  • Supported testing and user experience analysis, developing an understanding of how technical delivery connects to business and user needs.
2022 — 2024

Loads of Traffic (LoT)

Affiliate Executive & Software Developer

Combined affiliate/commercial work with Python automation, gaining early exposure to marketing and the eGaming sector.

  • Developed Python automation tools to streamline outreach and web scraping, removing manual processes and improving efficiency.
  • Optimised workflows by eliminating unnecessary stages, reducing operational complexity and improving scalability.
  • Applied data-driven analysis to improve affiliate performance and support decision-making.

05 / Skills
The toolkit.

8 categories
+ certified

Languages

PythonTypeScriptJavaScriptDartJavaSQLVBA

Web

ReactNext.jsTailwind CSSHTMLCSS

Mobile

FlutterDartCross-platform mobile developmentFirebase

Backend & APIs

Node.jsREST APIsMySQLPostgreSQLSupabaseStrapi CMSUmbraco

Data & AI

pandasData preprocessingFeature engineeringStatistical testingClassification modelsModel evaluationNLP (BERT / FinBERT)SHAPPermutation importanceApplied AILLM experimentation

Cloud & DevOps

Google Cloud Platform (GCP)VercelGit / GitHubCI/CD (GitHub Actions)Linux

Product & process

Certified Scrum MasterAgile / ScrumRequirements gatheringMVP scopingHCI methodsUser-centred design

Tools

JiraFigmaPostmanVS CodeCursorClaudeGitHub Copilot

06 / Education
On paper.

2023 — 2026
Loughborough University

BSc Computer Science

Loughborough University · 2023 — 2026
77.1%
Final degree mark · First-Class Honours
Embedded Systems Programming98%
Team Projects91%
Computer Science Project83%
Data Mining & Machine Learning79%
Object-Oriented Programming87%
✶
Entrepreneur of the YearNominated — Loughborough University, 2023.

07 / Writing
Some notes.

03 recent
all writing →
001   ESSAY

One source of truth for my CV and this website

Why I moved my career details into a single structured file that feeds both my LaTeX CV and this Next.js site, so they can never quietly drift apart.

Jun 2026 · 1 min readRead →
002   ESSAY

Diagnosing FinBERT's neutral-class blind spot

How an explainability-guided method traced a financial sentiment model's weakest class back to a single failure mode, and how a small, targeted fine-tune fixed it.

Jun 2026 · 2 min readRead →
003   ESSAY

Notes on owning Flutter architecture at MezzeData

What I have learned leading the Flutter side of a production app: layering for change, turning static PDFs into living content, and keeping a part-time codebase maintainable.

May 2026 · 2 min readRead →
08 / Contact

Let’s build
something good.

github / dominichubblelinkedin / dominichubblecv / download PDFemail / @gmail.com