Callum Cummins

Callum Cummins

Full-stack Software Engineer shipping C#/.NET, React/TypeScript and Python in production - features, performance work and root-cause fixes on business-critical platforms, plus LLM tooling that cut investigation time from hours to minutes.

United Kingdom · Open to interesting conversations

View workGet in touchDownload CV

Experience

Work experience

Where the code ships to production and real people depend on it.

  1. Junior Business Systems Solutions AI Engineer

    bet365 · Manchester, UK

    Oct 2025 - Present

    Full-stack engineer across C#/.NET, React/TypeScript and Python, shipping features, performance fixes and AI tooling for internal business platforms.

    • Own features end to end on a .NET 8 / React (RTK Query) trading operations platform: new API endpoints and stored procedures, repository layer, and the UI panels, state and tests (26-case comparer suite) that sit on top of them
    • Profiled and fixed hot paths in a high-traffic events API - Redis cache stampedes, a missing index, a ~5s DataTable intermediate and serializer options being rebuilt ~111k times per request - and diagnosed connection pool exhaustion
    • Root-caused a data-persistence bug spanning five separate faults (string comparison, stale client cache, discarded PUT response, client-side business rule) and moved the rule server-side behind a new endpoint
    • Built audit logging for a regulatory compliance platform using an EF Core SaveChangesInterceptor with write-time diffs, plus a generic incremental Postgres to BigQuery sync orchestrated with Hangfire
    • Reviewed the architecture of an LLM-based federated search proof of concept, flagging prompt injection, token storage and an undesigned streaming path before build
    • Within three months of joining, solely designed and built the Next.js/TypeScript frontend for a compliance investigation platform used across multiple regions, from blank repo to production - cutting case lookup from hours to ~8 minutes - and co-developed its Python / Azure OpenAI reporting workflow (10x+ analyst throughput)

About

About me

I'm a full-stack engineer at bet365, working across C#/.NET APIs, React/TypeScript frontends and Python tooling on the internal platforms that trading and compliance teams use every day.

Most of my time goes on the work between the feature and the production incident: profiling an API until a serializer being rebuilt 111,000 times a request shows up, tracing one "it doesn't save" report back to five separate faults, or catching prompt injection missing from an AI proof of concept's risk register before it gets built.

Outside work I keep building - from fine-tuning BERT for phishing detection at 97.7% validation accuracy to a voice-controlled AI home assistant - always with the goal of turning research-grade AI into software people can actually use.

Before bet365 I read Computer Science at Keele, graduating with First Class Honours and an 80% dissertation: a full-stack fitness tracker with predictive ML, built iteratively with real users.

Full-stack at bet365

.NET 8 APIs, React/TypeScript UIs and Python in production

Performance & root-cause

Cache stampedes, missing indexes, multi-fault persistence bugs

AI in production

LLM reporting workflows and architecture review of AI systems

First Class Honours

BSc Computer Science from Keele University

Education

  1. BSc Computer Science

    Keele University · First Class Honours

    Sep 2021 - July 2025
    Software Development ManagementAdvanced Databases and ApplicationsCommunications and NetworksComputational and Artificial Intelligence

    Dissertation: First Class grade of 80%. Gym Goer Web App: Developed a full-stack fitness tracking application with predictive ML capabilities, achieving 4.2/5 user satisfaction and 90% recommendation rate.

  2. Level 3 Extended Diploma in IT

    Coleg Cambria · Merit, Merit, Distinction (MMD)

    Sep 2019 - July 2021
    Computer SystemsComputer NetworksCyber Security3D AnimationGame DesignIT Project Management

Certifications

  • Exploring Artificial Intelligence: Use Cases and Applications

    AWS Training & Certification

    View credential
    Aug 2025

Beyond code

  • Bouldering - university competitions; problem-solving on the wall
  • Tutoring - volunteer programming tutor for local students
  • Fitness - regular training; consistency over intensity

Skills

Skills & expertise

Languages

PythonJavaScriptTypeScriptC# (.NET)PHPSQLHTML/CSS

Frameworks

ReactNext.jsViteTailwind CSSASP.NET CoreFlaskFastAPIPySide6 (Qt)

Libraries

Entity Framework CoreDapperHangfireRedux Toolkit / RTK Querystyled-componentsAnt Designscikit-learnPyTorchjQueryTkinter

I use C#/.NET, TypeScript and Python professionally - APIs, stored procedures and React frontends on the same ticket - with a habit of profiling before guessing.

Projects

Featured projects

AI experiments, full-stack apps, and coursework that got out of hand.

AI/ML

Jarvis - AI Home Assistant

Voice-controlled AI home assistant built on Microsoft Semantic Kernel, shipping a Rich terminal UI and a PySide6 desktop app with a full voice pipeline. Work in progress. "Is that all, sir?"

PythonSemantic KernelOllamaOpenAI APIPySide6 (Qt)faster-whisperopenWakeWordedge-tts
  • "Hey Jarvis" wake-word pipeline: speech-to-text, agent, text-to-speech, then a natural follow-up loop
  • Dual-provider: local Ollama and cloud OpenAI run simultaneously, switchable live by voice with isolated per-provider history
  • Function-calling plugins for Chrome browser control, Tavily web search, notes/todos, screenshots, and image generation
3D-Globe
LiveWeb DevelopmentAI/ML

3D-Globe

An interactive 3D globe where clicking any country opens AI-generated facts, landmarks, and languages on demand - combining geospatial processing, 3D rendering, and AI integration.

Next.jsTypeScriptThree.jsReact Three FiberD3-GeoZustandSWRGoogle Maps APIOpenAI APITailwind CSSJest
  • Interactive 3D globe with smooth spin/zoom inertia and camera tweens
  • Country detection for all countries using ISO3 codes and GeoJSON
  • On-demand AI facts panel with regenerate, backed by an in-memory LRU cache
CodeLive demo
Gym Goer Web App - Dissertation Project
AI/MLWeb Development

Gym Goer Web App - Dissertation Project

Full-stack fitness tracking application with predictive ML capabilities using custom AI models for exercise progress prediction.

HTML/CSSPHPPHPmlJavaScriptTailwind CSSMySQLChart.js
  • Custom AI model using SVR/Linear regression
  • Google Maps and OpenWeather API integration as well as own RESTful API for saving running routes
  • Agile development with user feedback iterations
Code
Phishing Email Detector
AI/MLData Analysis

Phishing Email Detector

Advanced email security application using BERT for phishing detection with high accuracy classification.

PythonPyTorchBERTTkinterPandasHugging Face
  • Fine-tuned BERT model on CUDA GPUs
  • 97.7% validation accuracy (0.98 precision, 0.97 recall) on 3.7K+ samples
  • OCR and clipboard input support
Code

More projects, contributions, and code samples on GitHub.

Contact

Get in touch

Got a project, a question or just want to chat about tech? Let's talk about it.

Email

callumcummins002@gmail.com

GitHub

CallumC28

LinkedIn

callum-cummins