The Bellini Line · Est. London

Harvey
Bellini.

Software engineer at Kraken. Based in London.

I build backend systems at Kraken: faster databases, trustworthy data and tools for running production work. Take the seat beside me to see what I’ve built, how it works and why the decisions matter.

Selected engineering / Kraken

Start the journey View all work

Python / Django / PostgreSQL / Kubernetes
Currently building at Kraken Technologies

BACKEND SYSTEMS. PRACTICAL RESULTS.Take the seat beside me.

01—07 / Work, study & interests

From energy systems to VR, football and the kitchen.

Systems

01 / Software Engineer I / Energy technology

Kraken.

Backend engineering for electricity settlement: large datasets, background processing and operational tools.

I work on Link, Kraken’s electricity settlement platform, using Python, Django and PostgreSQL. My work ranges from fixing slow queries and reclaiming database storage to building job execution and sign-in workflows.

1.43 TBof PostgreSQL storage reclaimed, approximately

Making large datasets usableRemoved timeouts from a view spanning approximately 26 million records. A separate message-query benchmark fell from about 63 seconds to 16.5 ms.Read the case study
Operational commands, from request to resultBuilt a permissioned Kubernetes Jobs workflow with tracked state, captured output, duplicate-execution guards and reconciliation.Read the case study
Fixing a queue consumer’s database bottleneckReplaced per-meter lookups with batches of 1,000. For 100,000 meter points, the query model falls from 200,001 queries to 201.Read the case study
One sign-in, two organisationsImplemented and tested OIDC sign-in with automatic tenant routing, then added default UI protection and signed access to a related service.Read the case study

More about my work at Kraken

Experiences

02 / Immersive Learning / VirtualSpeech

VirtualSpeech.

Django APIs and document ingestion behind AI-powered roleplay in VR.

I developed Python and Django REST APIs with VR developers, integrated document ingestion into roleplay scenarios, and worked on AWS infrastructure and SQL databases.

Document → dialoguebringing external material into roleplay scenarios

APIs for AI roleplayBuilt Python and Django REST APIs with VR developers for AI-driven roleplay applications.
Documents into scenariosIntegrated external documents through file ingestion and processing, with supporting JavaScript, HTML and CSS features.
Behind the VR experienceManaged AWS infrastructure and optimised SQL databases for the platform.

Read the project story

SystemsCommunityTwo interests

03 / MSci Computer Science / University of Exeter

My master’s degree.

MSci Computer Science at the University of Exeter, graduating with a 2:1 in 2024.

I completed an MSci in Computer Science at the University of Exeter, graduating with a 2:1 in July 2024.

MSci · 2:1Computer Science, University of Exeter

Building alongside the degreeCreated a Python and Three.js football visualisation with Manchester City.Read the case study
Leading a project teamLed six people to deliver a Django application for CoLab Exeter.Read the case study
Teaching PythonDesigned and taught a ten-week programming fundamentals course for university students.Read the case study

About my degree

ExperiencesOff DutyTwo interests

04 / F-Vision / Manchester City

F-Vision.

A ten-week project turning club-provided data into an interactive 3D football tool.

I built F-Vision with Python and Three.js during a ten-week project using club-provided data, and demonstrated it at the Etihad Stadium.

10 weeksfrom club-provided data to an Etihad demonstration

Python meets Three.jsBuilt an internal 3D football visualisation using club-provided data during a ten-week project.
Presented at the EtihadDemonstrated F-Vision to a judging panel including the club’s Head of Technology.

Read the project story

SystemsCommunityTwo interests

05 / Community Software / CoLab

CoLab Exeter.

Leading a team of six to build and deploy a Django mental-health application for an Exeter charity.

I led a team of six, translated a charity’s requirements into a Django mental-health application, and delivered its cloud deployment and migration.

6-person teamrequirements, application delivery and cloud deployment

A team of sixLed delivery of a Django mental-health application for an Exeter charity.
A move without interruptionMigrated the application from Google Cloud to DigitalOcean with zero downtime.

Read the project story

Community

06 / Python Workshops / Exeter Entrepreneurs

Teaching Python.

A ten-week introduction to Python, designed and taught for university students.

I designed and taught a ten-week Python fundamentals course for university students.

10 weeksof Python fundamentals, designed and taught

Designed and taughtCreated a ten-week Python fundamentals course for university students.
Sharing the starting pointWorkshop Leader for Exeter Entrepreneurs, September–December 2022.

About the course

Off Duty

07 / Recipes / Off Duty

The kitchen.

Focaccia, banana bread and tiramisu. Three recipes with ingredients beside each step.

I keep a small recipe collection with ingredients and instructions together, ready to cook from or print.

Something to sharea small collection for the kitchen

Three recipes to come back toFocaccia, chocolate chip banana bread and tiramisu, with ingredients alongside each step.
Ready to cook fromSearch the collection, open an individual recipe or use its print layout.

About the collection Browse the recipes →

The engineering log / Selected work at Kraken

Behind the service.
Inside the systems.

My production work is in energy technology. The problems are about performance under load, the integrity of data and knowing what happened when a process fails.

01Performance / Kraken Technologies

Make the database do less work.

63 s → 16.5 msRecorded SQL execution benchmark

The problem: An operational message view combined and sorted every matching record before taking a single page of results.
What I changed: I bounded each ordered input before combining the results, and avoided the union when only one message direction was needed.

Unbounded inputs
Limit before combining
Read one page

The engineering principle

Find the unnecessary work before adding more infrastructure. For data-heavy systems, the shape of a query can matter more than the size of a server.

These are SQL execution timings, not whole-page latency. Separately, I removed 11 redundant indexes and reclaimed approximately 1.43 TB across two production estates.

Read the full case study

02Correctness / Kraken Technologies

Keep bad data out of settlement.

≈483k / monthInvalid updates blocked across two clients

The problem: Billing validation and industry submission run asynchronously. A reading could leave the system before a separate process quarantined it.
What I changed: I added reading-validation safeguards and persisted suspect-reading state so a separate submission worker could see it. A configurable delay also gives asynchronous billing time to run.

Validate the reading
Persist shared state
Gate submission

The engineering principle

Correctness has to survive the boundary between processes. A local flag is not enough when another worker makes the next decision.

The monthly count is the validation result recorded in my CV. It does not measure the separate submission-delay or migration-repair changes.

Read the full case study

03Controlled operations / Kraken Technologies

Give production work a visible lifecycle.

Request → resultPermissioned execution with captured output

The problem: Launching a background command is only part of the job. Competing requests must not run equivalent work concurrently, and failed work must not stay marked as running.
What I changed: I built the Django models, execution logic, Kubernetes integration and UI: explicit permissions, ordered row locks, captured output and reconciliation with the actual Job state.

Permissioned request
Claim + Kubernetes Job
Output + reconciliation

The engineering principle

Make it possible to explain who can run work, what happened and how failures are recovered. Operational control belongs in the system itself.

Delivered across four implementation stages. The source supports the workflow and concurrency safeguards; it does not quantify adoption or time saved.

Read the full case study

Find a project or follow the line

Work & interests.

All contributions ↗

SystemsBackend engineering, scale & reliability
ExperiencesInteractive products & visualisation
CommunityLearning, leadership & teaching
Off DutyFootball, food & personal projects

01
KrakenKraken Technologies
02
VirtualSpeechVirtualSpeech
03
My master’s degreeUniversity of Exeter
04
F-VisionManchester City FC
05
CoLab ExeterCoLab Exeter
06
Teaching PythonExeter Entrepreneurs
07
The kitchenA personal collection

Behind the line

Software, football
and a little baking.

I’m based in London and work on Link at Kraken Technologies. Before Kraken, I built backend services for AI roleplay at VirtualSpeech. I graduated from Exeter with an MSci in Computer Science in 2024.

Along the way, I’ve built a 3D football tool with Manchester City, led a charity software project and taught Python. Outside work, I’m into football, motorsport, consumer electronics and gaming. The kitchen is where the recipe collection comes in.

Download my CV ↓

There’s always another connection.

Let’s talk

Interesting problems. New ideas. A good conversation.

harveybellini@gmail.com

HarveyBellini.

Kraken.

VirtualSpeech.

My master’s degree.

F-Vision.

CoLab Exeter.

Teaching Python.

The kitchen.

Harvey Bellini.

Behind the service.Inside the systems.

Make the database do less work.

Keep bad data out of settlement.

Give production work a visible lifecycle.

Work & interests.

Software, footballand a little baking.

Harvey
Bellini.

Behind the service.
Inside the systems.

Software, football
and a little baking.