Skip to content
robinSenior Software Engineer - Applied AI

Robin White

Senior Software Engineer — AI Infrastructure & Systems
North London, UK / New York, USA
07557525742 | robinojw@gmail.com | robinw.co.uk
Summary

Senior Software Engineer specialising in AI agent infrastructure, distributed systems, and developer tooling. At Peloton, designed and shipped the ChatGPT platform integration (launched at OpenAI Dev Day), built production MCP servers and generative UI services, and architected a hybrid vector search system completing full-catalogue queries in under 30ms. Built large-scale change agents using LLMs and RAG to operate across a 9-million-line monorepo. Deep experience with agent orchestration patterns, execution environments, CI/CD pipeline design, and cloud infrastructure (Cloudflare Workers, Kubernetes, serverless). Currently building agentic developer tooling and multi-agent research pipelines independently.

Work Experience
Peloton — Senior Software Engineer
Aug 2024 – Present
New York, USA
  • •Designed and implemented the Peloton ChatGPT app using the OpenAI Apps SDK, launched by Sam Altman at OpenAI Dev Day. Built the execution and rendering infrastructure: a Preact-based generative UI system delivering interactive widgets in 45–70KB payloads with sub-500ms tool responses inside ChatGPT's iframe sandbox.
  • •Architected and shipped production MCP servers enabling third-party platform integrations, designing tool schemas, auth boundaries, and the orchestration layer between model requests and internal APIs.
  • •Built a hybrid vector and keyword search API in Go indexing Peloton's class catalogue with caption embeddings. Tuned retrieval to complete full-catalogue searches in under 100ms on Cloudflare infrastructure.
  • •Designed Postgres and serverless database infrastructure for new AI-facing services, including connection pooling (Hyperdrive), caching layers, and edge execution patterns.
  • •Owned on-call rotations for production web services via PagerDuty, leading incident triage, resolution, and blameless postmortems. Monitored service health through Datadog RUM, Vercel Alerts, and Core Web Vitals dashboards, with automated smoke and regression test suites (Playwright, Jest) gating every deployment.
  • •Led migration of production web properties from Netlify to Vercel & Cloudflare, managing cutover strategy, DNS transitions, build pipeline reconfiguration, and performance validation across applications serving millions of users.
  • •Led cross-application platform initiatives spanning multiple teams and services, owning technical direction and execution timelines.
Peloton — Platform Software Engineer III
Jan 2024 – Aug 2024
New York, USA
  • •Designed and deployed large-scale change AI agents using OpenAI and Anthropic models with RAG tooling to automate migrations, dependency updates, and circular dependency removal across a 9-million-line monorepo containing 23+ applications.
  • •Led platform-wide web architecture redesign, establishing patterns for server-side rendering, edge caching, and application isolation across the organisation's web properties.
  • •Built performance monitoring infrastructure for large-scale web applications using Datadog RUM, Core Web Vitals tracking, and Vercel observability tooling. Established alerting thresholds, regression detection, and end-to-end test coverage with Playwright to catch degradations before they reached production.
  • •Managed cloud provider migration of 50+ applications, including capacity planning, cost modelling, and contract negotiation, saving over $260,000 in annual infrastructure costs.
Peloton — Software Engineer III (Tech Lead)
Mar 2023 – Jan 2024
London, UK
  • •Technical lead on a system to make 40,000+ classes publicly discoverable, designing the ISR pipeline to manage over 200,000 statically generated pages with on-demand invalidation while protecting internal infrastructure from traffic spikes.
  • •Built a unit-testing and coverage pipeline that reduced execution time by 56% and coverage calculation by 74%, improving developer feedback loops across the organisation.
Peloton — Software Engineer II
Nov 2021 – Mar 2023
London, UK
  • •Designed and implemented a CI build system that reduced pipeline cost by 93% per PR and median runtime by 60%, using dynamic job scheduling in GitHub Actions.
  • •Architected Kotlin/Spring microservices during migration from a Python monolith, designing service boundaries, API contracts, and deployment strategies.
Phantom — Developer
Aug 2020 – Nov 2021
London, UK
  • •Led development on projects for Google, Sony, the Financial Times, and Zendesk, working across Angular, React, Python, and Bazel-based systems.
  • •Built three products in the Google monorepo using Bazel, Protocol Buffers, and Google Cloud infrastructure.
  • •Created internal and open-source developer tooling to maintain velocity and code quality across teams.
Zebra Technologies — UX Developer (Internship)
2018 – 2019
London, UK
  • •Built a VR application in Unity/C# for the NFL, debuted at the company's annual sales conference.
  • •Developed and maintained Android applications for the company's industry-focused mobile devices.
Independent Projects
DJ — Agentic Orchestration TUI
  • •Building an agentic orchestration terminal interface for managing multi-model coding loops, focused on minimising harness overhead and maximising tokens-per-second throughput.
  • •Analysing end-to-end latency in agentic execution: tool-call round-trips, context window management, and the architectural implications of inference speed scaling from 75 to 8,000+ TPS.
Parameter Golf — Multi-Agent Research Pipeline
  • •Competing in OpenAI's Parameter Golf challenge. Built a multi-agent research system that conducts literature search, grades papers by relevance, generates novel compression hypotheses, and validates them by spinning up H100 GPU instances on RunPod and locally with MLX.
  • •Architecture: orchestrator agent dispatches specialised sub-agents for research, experimentation, and evaluation, with automated feedback loops.
Outside — Agentic Coaching Platform (trainingoutside.com)
  • •Designing and building an agentic coaching platform on Cloudflare Workers AI, Next.js, and Capacitor.
  • •Implemented tool-use agents enabling athletes to manage training plans through natural language and conduct research via a retrieval-augmented interface pulling from coaching and sports science literature.
Technical Skills
Languages: Go, TypeScript, Python, Kotlin, C#
AI/Agent Infrastructure: OpenAI APIs (Apps SDK, Assistants, Chat Completions), Anthropic APIs, MCP servers, RAG pipelines, vector search (embeddings, hybrid retrieval), prompt engineering, agentic orchestration
Systems & Infrastructure: Cloudflare (Workers, KV, Hyperdrive, R2), Kubernetes, AWS, GitHub Actions, CI/CD pipeline design, serverless architecture
Observability & Reliability: PagerDuty, Datadog RUM, Vercel Alerts, Core Web Vitals, Playwright, Jest, incident response
Backend: PostgreSQL, edge databases, microservice architecture (Spring/Kotlin), REST API design
Frontend: React, Next.js, Preact, Electron, Node.js
Build Systems & Tooling: Monorepo CI/CD at scale, Bazel, dynamic pipeline orchestration, developer tooling
Education
University of East Anglia
BSc Computer Science with a Year in Industry — 2:1
West Suffolk College
Extended Diploma in Computer Science — Triple Distinction Star (A*, A*, A*)