Joe Dempsey wearing sunglasses against an orange background

Joe Dempsey - Innovation focused technologist based in London

Products are built by great teams. That’s where the work starts.

Joe Dempsey is a London-based engineering leader with 15+ years across consulting, finance, travel and air traffic control. Today he leads GenAI engineering teams at a global financial-data company, where the focus is turning AI ideas into products people rely on.

The loop is simple: identify > quickly test > scale > optimise. The job is building teams that can run it fast and safely, together.

2026 Focus: Scaling GenAI

  • Agentic platforms. Production-grade foundations covering model and AI gateways, memory, observability and evaluation. Models and tools are reached only through the gateways, so product teams get auth, metering and tracing for free and can ship AI features without reinventing the platform.
  • A better way to build. A shared, AI-first toolkit for every product role: PMs, designers, engineers and QA. That means AI-ready repos, common standards, training and sensible guardrails. The goal is to make the right path the easy path for everyone. Thin harness, domain knowledge in shareable skills.
  • Trust by design. Layered evaluation for multi-step agent workflows, combining automated judges, human review and an honest view of cost versus quality. We write the exam first, from real failures, and hard gates can veto a release. No single check is perfect, so the layers back each other up.
  • Partnerships. Partnering across the major AI ecosystems and hyperscalers through open protocols (MCP, A2A) and APIs, so trusted data shows up wherever customers already are. Build once, reach every channel.
  • A healthy innovation cycle. Clear paths from proof of concept to production. Move fast where it’s reversible, gate carefully where it isn’t. Ideas that don’t land are retired on purpose, which frees the team for the next bet.

2025 Focus: Innovation and problem-solving

  • Early-stage ideas. Leading engineering on new concepts and testing whether each one is desirable, feasible and viable before choosing the right architecture.
  • Proofs of concept. Building fast POCs across AI, GenAI and blockchain to separate real opportunity from hype.
  • Global Hackathon. All ideas revolving around the power of AI and its potential to transform businesses. The three key themes challenged participants to bring their best:
    • Hack my Workflow
    • GenAI-Powered Customer Solutions
    • Moonshot - Bending the Curve
  • https://hbr.org/2025/03/how-a-legacy-financial-institution-went-all-in-on-gen-ai?linkId=13683957

2023–2024 Focus: New ventures and launch

  • Startup appraisal. Assessing startups and emerging technology for fit, risk and potential.
  • MVP to launch. Taking new ventures from MVP to launch, including a full SDLC, technical compliance and the foundations to scale.
  • Broad perspective. Work across many clients and sectors gave the team a broad outlook, a knack for explaining technical ideas to non-specialists, and strong stakeholder relationships.

How I work

  • Direction over control. Set a clear goal, give context, set checkpoints, then get out of the way.
  • Scale through people. Every process should outlive the person who started it. Deputies, playbooks and self-service beat bottlenecks.
  • Bring everyone along. Change only sticks when the whole organisation moves, not just early adopters. Coaching and enablement are part of the job.
  • Grow people, not just products. Stretch roles, honest feedback and room to own outcomes build leaders at every level.
  • Make it safe to fail fast. Small bets, quick learning and no blame when an experiment doesn’t land.
  • Partner widely. The best outcomes come from close work with product, risk, security and external partners.
  • Balance the pull. Innovation and delivery compete for the same people. Make the trade-off explicit.

What i’m exploring

  • Self-improving agents and agentic “software factories”
  • Confidential AI and secure inference over sensitive data
  • Context engines and AI-ready codebases
  • Sandboxed, cloud-run coding agents
  • Model routing and open-weight models for engineering
  • Blockchain, and where it meets agentic systems

Some of this gets shared through talks on platforming, agentic development and the future of developer experience.

Mindset

Learning never stops, and neuroplasticity proves it. Strong teams treat curiosity as a habit: they go deep fast, challenge each other, lean on experts, and turn what they learn into something real.

Beyond work

Based in London, dad of two. Home life is a constant lesson in prioritisation and teamwork.

Connect with me on LinkedIn.