~/jeeson-johnson

$ whoami

I build intelligent systems — and the platforms they run on.

> Senior AI & Software Engineer · IBM · London, United Kingdom

Currently the technical lead for Cloud Pak for Integration’s AI agents — MCP servers, RAG pipelines, LangGraph multi-agent systems, built on a foundation of Go, Kubernetes. On the side, I’m co-founder & CTO of mealzy.io.

// experience

Jun 2025 — Present

London, UK · Hybrid

Senior AI & Software Engineer · IBM

Technical lead for the CP4I AI agents — owning architecture from framework selection down to implementation detail.

  • Designed and built a stateful LangGraph multi-agent system in Python for Cloud Pak for Integration
  • Prototyped the MCP server and LangGraph agents that convinced leadership to prioritise CP4I's first AI-enabled capabilities
  • Engineered an LLM-as-a-judge E2E evaluation framework for non-deterministic agent outputs — dual-mode execution (mocked for fast CI, live MCP for full E2E), promoted at VP level and used across departments
  • Implemented RAG pipelines over a curated knowledge store and ReAct-style agents; benchmarked LLMs on cost vs. performance to select production models
  • Replaced validation webhooks with Kubernetes CEL to support AKS — adopted by the Architectural Board as departmental best practice
  • Mentored junior engineers and host education sessions on MCP development and LangGraph agents

Recognised with two Outstanding Technical Achievement Awards, and a place at IBM Tech 2026 — ranked in the top 0.5% of IBM's global workforce.

Sep 2022 — Jun 2025

Winchester, UK

Software Engineer · IBM

Cloud-native platform engineering across the CP4I stack — Node.js APIs, Go operators and React UIs, shipped end-to-end.

  • Led cross-repo projects from stakeholder requirements to delivery: Node.js APIs, OpenShift operators in Go, and React UI changes, with unit, integration and E2E coverage
  • Cut E2E test cycles from 30+ minutes to 8 by architecting a Docker-based, cross-browser visual-diff CI pipeline — presented department-wide as a model for testing
  • Reduced bulk instance provisioning time by 73% by redesigning a synchronous creation flow into an asynchronous, dependency-aware algorithm
  • Architected CP4I's Call Home telemetry pipeline in Go (Segment + Amplitude) with zero performance regressions — a pattern reused by other teams
  • Built the CP4I Integration Assembly Canvas, a drag-and-drop interface for multi-product infrastructure
  • Built a VS Code extension automating OpenShift cluster connection and configuration for the wider engineering team

Apr 2020 — Apr 2021

Winchester, UK

Software Engineer · Internship · IBM

Core feature development for Cloud Pak for Integration and IBM API Connect.

  • Cut API load by 80% through fuzzy contextual search and caching optimisations
  • Implemented end-to-end enterprise features, including a secure JWT-based user management system
  • Led delivery of asynchronous backup generation, contributing fixes upstream to open-source libraries

// projects

live

Co-founder & CTO

mealzy.io

One QR. Everything customers need.

A QR-first customer platform for restaurants, cafés and local businesses. One scannable code bundles a venue's menu, reviews, booking and social links into a single hub — backed by a digital stamp-card loyalty programme, opt-in email and SMS outreach, scan analytics, and a drag-and-drop website builder for commercial clients. The pitch: owned marketing channels and repeat visits, with no new hardware and no app to install.

  • Next.js
  • TypeScript
  • Node.js
  • Vercel AI SDK
  • PostgreSQL · Prisma
  • Stripe
  • Cloudflare
  • Google Cloud

// capabilities

AI Engineering

  • LLM agents · ReAct, LangGraph
  • MCP server development
  • RAG system design · Milvus
  • Prompt & context engineering
  • LLM-as-a-judge evaluation
  • Agentic architecture design
  • Model cost/performance analysis
  • Vector embedding systems
  • Python

Platform & Cloud

  • Go · Kubernetes operators
  • OpenShift · Docker
  • Node.js · Express · PostgreSQL
  • AWS · IBM Cloud · GCP
  • REST APIs · Microservices
  • CI/CD · Jenkins, GitHub
  • Async orchestration engines
  • Telemetry · Segment, Amplitude
  • L3 production support

Product Engineering

  • React · Next.js · TypeScript
  • Complex canvas interfaces
  • VS Code extensions
  • E2E testing · Cypress, Jest
  • Visual diff CI pipelines
  • System design · TDD
  • AI tooling · Claude Code, Codex, Cursor
  • Developer tooling
  • Technical writing & talks

// recognition

// patent

Patent pending · five defensive publications

A U.S. patent application filed on behalf of IBM for vector-embedding innovations in AI support workflows, plus five defensive publications across AI, security and automation.

// award

Outstanding Technical Achievement Award ×2

Awarded twice by IBM — including for driving the architectural decisions behind the CP4I AI agents.

// award

Entrepreneur Award

For co-building a cross-department E2E testing framework for agentic AI — LLM-as-a-judge evaluation of unit tests, integration tests and cost metrics, running against live clusters or mocked data.

// recognition

IBM Tech 2026 — San Diego

Nominated for work on the CP4I AI agents — ranked in the top 0.5% of IBM's global workforce for exceptional contributions.

// certification

AWS Certified Cloud Practitioner

Issued by Amazon Web Services — cloud fluency and foundational knowledge of the AWS service landscape.

// certification

IBM OIC Initiative Leadership

For leading multiple Open Innovation Community initiatives — hosting teaching sessions and talks within IBM's Front-End Development community.

// accreditation

Developer Profession, Levels 1 & 2

IBM's engineering profession accreditation, held through to Level 2 — Experienced/Senior Engineer.

// journal

Have a hard problem? Let’s talk.