← Explore the townGet in touch ↗
Pixel-art portrait of Jack Winter

Jack Winter.

Senior Software Engineer, Frontend · London

About

I’m a senior software engineer with 8+ years of experience building customer-facing products with React and TypeScript across sports betting, gaming, and fintech.

I’ve led greenfield builds and major frontend transformations, owning architecture and delivery from the initial problem through to production. I specialise in real-time interfaces and performance optimisation, and build open-source tools for AI coding agents using MCP, embeddings, and semantic code retrieval.

Projects

WinterBot

Winter Town’s paths, buildings and gardensTry the demo

How I built Winter Town ↗

This portfolio, built with Astro, Preact, and TypeScript.

My contribution
Built the town, movement and navigation in Preact, with an Astro portfolio sharing the same content.
Evidence
Browser tests cover movement, navigation, saved progress, and the JavaScript-free portfolio. Read more ↗
Behind the project
Problem
Make a portfolio worth exploring while keeping projects and contact details easy to reach.
Tradeoff
Exploration is optional: direct links and keyboard shortcuts reach every section.
  • Astro
  • Preact
  • TypeScript
  • Playwright

brain-cache

Recorded source · 34a4ebbassembleContext()
Find code: “budget”src/services/tokenCounter.ts
Keep complete chunks within a token budget.
Try the demo

Helps coding agents find relevant functions without reading entire files, using a local code index and MCP.

  • Splits code around functions and classes to preserve useful context.
  • Reuses embeddings and indexes changed content instead of rebuilding after every edit.
  • Owned architecture, search, CLI, assistant integration and npm release.
Behind the project
Problem
Coding agents need relevant source without reading the whole repository for every question.
Tradeoff
Smaller results can omit important evidence. Local indexing needs model setup and machine resources; retrieved snippets may still go to the assistant’s cloud model.
Evaluation
Tests cover indexing, retrieval, context assembly and packaging. Context-size estimates are not evidence of better task outcomes or whole-session cost savings. Read the evidence ↗
  • TypeScript
  • Node.js
  • MCP
  • LanceDB
  • Ollama
  • tree-sitter

pancake-stack

Walking regression replayA frame arrived before the click
Negative elapsed time → NaN positionClamp elapsed time → valid movement
Replay the failure and its fix.
Try the demo

Reusable workflows for coding agents, built to address rushed changes, unnecessary complexity and unverified claims of success.

  • Designed task-specific investigation, implementation, review and verification workflows.
  • Built a separate evaluator using known defects and correct controls to check missed bugs and unsupported findings.
  • Rejected extra guidance when existing instructions caught the same targeted defects.
Behind the project
Problem
Fast AI-generated changes can leave more cleanup than useful code.
Tradeoff
The plugin supplies instructions, not an enforcement runtime. More process has a cost, and completed steps do not guarantee a correct result.
Evaluation
Independent review found that combining two seeded defects could reward a test that caught only one. The evaluator was corrected to test each defect separately. These bounded trials do not establish general productivity or quality gains. Read the evidence ↗
  • Codex
  • Claude Code
  • Agent Skills
  • Python
  • GitHub Actions

Experience

WINTER TOWN RAILWAY2017 → PRESENT
  1. Jun 2025 — Present · Remote

    KTO Group

    Senior Front-End Developer

    Sportsbook and online casino operator serving the Brazilian market.

    Read the KTO case study ↗

    • Led the transformation from a third-party hosted sportsbook to a custom React and TypeScript platform, owning architecture and core customer journeys within a business serving 1.4M active users and 5B bets placed.
    • Delivered 40% faster FCP, 25% faster LCP, 20% lower blocking time, 9.3% less transferred data, and 11× faster return navigation in performance testing.
    • Designed a modular sportsbook architecture across a pnpm/Turborepo monorepo spanning 27 packages and 22 modules, separating reusable UI, domain logic, data providers, and third-party integrations.
    • Built a real-time sports data pipeline with Web Workers, Socket.IO, and Zustand, moving high-frequency events and odds processing off the main thread.
  2. Jun 2023 — Jun 2025 · London

    GeoGaming Group

    Senior Front-End Developer

    Sportsbook and casino games provider developing a white-label solution.

    • Led frontend performance initiatives that reduced application load times by 30%.
    • Modernised bundling, code splitting, and lazy loading to reduce unnecessary client-side overhead.
    • Increased automated test coverage from 25% to 75%, strengthening unit, integration, and end-to-end testing with Jest and Cypress.
    • Built real-time betting interfaces with RxJS, WebSockets, and REST APIs.
    • Mentored junior engineers, led code reviews, and helped establish frontend engineering standards.
  3. Jul 2017 — Aug 2022 · Bristol

    SmartStream Technologies

    Senior Software Engineer

    Fintech developing cash and liquidity management software for major banks.

    • Led a greenfield Angular frontend that replaced a legacy system and directly enabled two high-profile deals.
    • Built internal tools with React and Node.js to streamline team workflows.
    • Developed a machine-learning prototype for transaction forecasting, improving forecast accuracy on historical banking data.
    • Interviewed candidates from junior to senior developer roles and worked with DevOps on critical system issues, including on-call support for 24/7 reliability.

Skills & education

Technical skills

Languages & frameworks

  • TypeScript
  • JavaScript
  • React
  • Node.js
  • Angular
  • Python
  • Java
  • HTML
  • CSS

Real-time & data

  • WebSockets
  • Socket.IO
  • RxJS
  • Zustand
  • TanStack
  • REST APIs
  • SQL
  • Web Workers

AI & developer tooling

  • LLMs
  • MCP
  • Embeddings
  • Vector search
  • Ollama
  • LanceDB
  • tree-sitter
  • Claude Code
  • Codex

Testing

  • Jest
  • Cypress
  • Vitest
  • Unit, integration, and end-to-end testing

Engineering & platform

  • pnpm
  • Turborepo
  • Storybook
  • Docker
  • AWS
  • CI/CD
  • Git

Education

BSc Computer Science · 2:1
University of the West of England

Sep 2013 — Apr 2017

Contact

For AI agents: each section is also plain markdown. Start at /llms.txt.