AI systems engineer · Toronto

Michael Skupien

I build AI systems that maximize human wellbeing. Today, AI agents for NHS primary care at Anima Health. Before that I ran engineering at Knockri making AI behavioural interview assessments. I started in robotics and co-founded two self-driving vehicle companies through Y Combinator (W16).

01

Current work

Lead, AI systems engineering · 2024 – now

Anima Health — AI agents for NHS primary care

I lead AI systems engineering for a care platform used by GP practices across the UK serving 14,000,000+ patients, turning consultation audio, clinical documents and patient calls into structured work and automations for care teams.

  • In-person consultation audioVoice note-takingStructured clinical note
  • Letters & clinical documentsContent extractionWorkflow routingClinical coding agentFully processed coded documents routed to the right person
  • Inbound patient phoneVoice agentUses the questionnaire agent: sends answers, gets the next questionsInbound patient appQuestionnaire agentRequests in the practice dashboard

Delivered inside Anima’s care platform, alongside the practice dashboard, triage, messaging and EHR integration (EMIS, SystmOne).

Anima AI systemsSimplified
Open source · MIT · TypeScriptAnima ADK — Build agents around your application’s state. ↗

I designed and built this open-source TypeScript library for text and voice agents that share one typed application state: context renderers choose what each agent sees, deterministic code and model calls mix in the same workflow, and durable pauses wait for a human before carrying on.

02

Previously

Founder & CEO · 2023

Bloomry — AI for mental health, pediatrics and education

A startup I founded after Knockri. Alongside 100+ customer interviews I built and pivoted between prototypes, including a video pipeline that cut a highlight reel of each child at daycare while redacting others, and 50 mental-health agents that held 3,000+ conversations.

Software engineer → ML engineer → engineering lead · 2020 – 2022

Knockri — AI interview assessment

I led a cross-functional team of eight taking AI hiring tools from research to production for 100,000+ candidates and hiring managers. I designed the patented scoring system behind it, which automates structured behavioral interview assessments to reduce racial and gender bias while keeping state-of-the-art correlation with on-the-job performance.

  1. Interview transcript
  2. Behavior classifier
  3. Skill scores
Behavioral scoring pipelineSimplified
Patent granted · US 2024/0256999 A1System and method for behavioral classification ↗

Co-inventor and lead engineer. Grounded in an I/O-psychology knowledge graph, with active learning and a trained-annotator program so every score is explainable.

03

Foundations — autonomy & safety-critical systems

2015 – 2018
04

Earlier projects

2012 – 2019
05

Press