Leonardo Aoun

Leonardo Aoun

Fractional AI Lead | Ex-Meta Research Engineer

High-impact technical leader with a background in Applied AI (Meta) and early-stage R&D (CTO, ML Lead). Scaling engineering organisations, bridging technical roadmaps with business strategy, and mentoring high-performing teams — fractional leadership for data-intensive AI startups.

About

High-impact technical leader with a background in Applied AI (Meta) and early-stage R&D as CTO and ML Lead. I specialise in scaling engineering organisations, bridging technical roadmaps with business strategy, and mentoring high-performing teams. Through Travelling Codesman I offer fractional leadership for data-intensive AI startups — cloud or on-prem — with the judgement to move research into systems that survive production.

Leonardo Aoun, smiling with arms crossed, wearing a navy jacket

Experience

  1. Freelance / Contracting

    2025 – present

    2025 – present

    Work experience

    Travelling Codesman Ltd

    2025 – present

    Website
    1. Freelance / Contracting

      2025 – present

      • Fractional leadership
      • Applied AI
      • R&D scaling

    London & remote

    Fractional and contracting work with startups and SMEs that need R&D teams to scale without losing delivery discipline — including agentic systems and production AI.

  2. ML Lead

    2024 – 2025

    2024 – 2025

    Work experience

    Noah Labs

    2024 – 2025

    Website
    1. ML Lead

      2024 – 2025

      • Acoustic AI
      • Mentoring
      • CI/CD

    Remote — Berlin

    Led ML at Noah Labs (Berlin-based; I worked remote) through the jump from research prototypes to systems that had to hold up in healthcare — process, mentoring, and acoustic AI in production.

  3. CTO & Co-founder

    2023 – 2024

    2023 – 2024

    Work experience

    Laces AI

    2023 – 2024

    Website
    1. CTO & Co-founder

      2023 – 2024

      • AWS
      • Next.js
      • Computer vision
      • Sales

    London

    Co-founded Laces AI and owned more than the technical platform. On the product side I architected cloud-native infrastructure on AWS and a Next.js web app delivering real-time computer-vision insights from models such as YOLO and StrongSORT.

    I was equally deep in sales and discovery: reaching out to clubs and coaches, learning their pain points, and turning those conversations into a sharper value proposition for video analysts and sporting directors. We also built agentic systems — before that language was everywhere — to automate parts of the sales and cold-outreach process so the team could spend time on the conversations that mattered.

  4. Machine Learning Engineer

    2020 – 2023

    2020 – 2023

    Work experience

    Meta AI Applied Research

    2020 – 2023

    Website
    1. Machine Learning Engineer

      2020 – 2023

      • Egocentric video
      • Computer vision
      • PyTorch Lightning

    London

    Three years on Meta AI Applied Research in London, on the Behavioural Computing team. The brief was anything human-centred that shows up in egocentric video — in our case, from the wearer’s point of view on the Ray-Ban Meta smart glasses programme.

    A large thread was automatic avatar generation and animation that stays true to the person: expressions and emotion cues vary across cultures, so a one-size-fits-all detector tends to misread people. We worked toward avatars and motion that respect how someone actually presents, rather than flattening everyone onto a single cultural template.

    Alongside that, the engineering work was the usual applied-research grind at scale — data pipelines over hundreds of thousands of video hours with PyTorch Lightning and internal distributed systems, cheaper inference and training (~70% compute reduction) without giving up accuracy, and getting research models into production-ready cloud infrastructure.

  5. Applied Science Intern

    Summer 2019

    Summer 2019

    Work experience

    Amazon

    Summer 2019

    Website
    1. Applied Science Intern

      Summer 2019

      • PyTorch3D
      • 3D retrieval

    Berlin

    Applied science internship with Amazon’s computer vision team in Berlin. I worked with PyTorch3D on a retrieval system that matched catalogue items by 3D shape — so similar geometry could surface even when texture or photos differed.

  6. Software Engineering Intern

    2018 – 2019

    Winter 2019

    Software Engineering Intern

    Summer 2018

    Work experience

    Facebook

    2018 – 2019

    Website
    1. Software Engineering Intern

      Winter 2019

      • Internal tools
      • Research collaboration
    2. Software Engineering Intern

      Summer 2018

      • Data privacy
      • Security

    London · New York

    Summer 2018 — London. Data privacy and security. I worked on systems that make sure user data is deleted when it should be — the kind of enforcement that only matters if it is reliable at scale. Related public write-up: Enforcing data deletion at Facebook scale (DELF).

    Winter 2019 — Facebook AI Research, Manhattan. I built toward an internal research-project discovery tool: a way for researchers and engineers to find relevant projects and collaborate across the organisation.

  7. Research Engineering Intern

    Summer 2017

    Summer 2017

    Work experience

    Adobe

    Summer 2017

    Website
    1. Research Engineering Intern

      Summer 2017

      • GPU clusters
      • Distributed training

    San Jose

    Research engineering internship with Adobe Research in San Jose. The work sat horizontally across the research organisation rather than inside one product team: standing up a GPU training cluster and benchmarking multi-node training options so labs across Adobe Research could train at scale.

Testimonials

  • Wilhelm, CTO · Terraspect

    Leonardo was a key partner during our rapid MVP iterations, ultimately building the stable task orchestration and agentic infrastructure we needed to move into production with confidence.

  • Science Machine

    Leonardo implemented a context optimisation strategy for our long-form agent discussions, unlocking significantly longer conversational coherence while maintaining high token efficiency.

  • Polaron

    Leonardo delivered enterprise-grade features and task orchestration for our workflows, including automated GPU provisioning and self-hosted infrastructure for our on-premise end-customers.