Building AI that
actually ships.

I'm Mason — an AI leader and researcher, building ML Systems specializing in Video and Multimodal Intelligence.
10+ years building teams and products in Critical Infrastructure Monitoring, and 5+ years building teams and products in Online Safety and Security, I build ML-driven teams and systems that go from whiteboard to production, from idea to reality.

See my work Get in touch
12+
years leading
technical teams
5+
years in online
safety & security
10+
years in critical
infrastructure
PhD
Computer Science
Dalhousie
About

The longer version

I build and lead teams that deliver production AI systems—focused on LLMs, agent-based architectures, and real-time, mission-critical ML platforms.

Most recently, as Director of Research and Machine Learning, I built and led a 12-person cross-functional organization responsible for designing and deploying enterprise-scale AI systems. I owned the full lifecycle—from defining ML strategy and system architecture to delivering production systems with measurable impact, including +35% model accuracy improvements, 45% infrastructure cost reduction, and reduced inference latency in live environments.

My current work centers on LLM-powered systems and autonomous agents:

I bring a hybrid background across machine learning, distributed systems, and cloud architecture, with deep experience building low-latency, high-reliability systems on AWS.

Alongside technical leadership, I’ve spent 12+ years building and mentoring teams across industry, academia, and hybrid environments. I’ve worked closely with researchers and engineers across disciplines—mathematics, physics, statistics, and computing science—and take pride in developing talent that has gone on to top-tier PhDs, leading tech companies, and venture-backed startups.

I also have extensive experience partnering with business stakeholders to translate ambiguous requirements into clear technical strategy and execution plans, ensuring that AI systems deliver real operational value—not just prototypes.

Outside of direct work, I’ve been actively involved in growing the local tech and research ecosystem in the Okanagan, supporting collaboration between industry and academia and helping develop the next generation of technical talent.

I value empathetic leadership, high ownership, and continuous learning—and I’m most motivated by complex, high-ambiguity problems where strong systems thinking and practical execution matter.

What I've been building

Recent production work across video intelligence, agentic systems, and generative AI.

🎬

Multimodal Video Pipelines

Automated ad safety, metadata labeling, theme detection, and captioning (with translation) for a family streaming platform.

🤖

Autonomous Agent Systems

Deployed agents to monitor infrastructure health and predict production risk. Also rolled out org-wide AI assistants for self-serve KPI access.

🔒

Content Safety & Moderation

Human-in-the-loop labeling workflows with active feedback loops. Content scoring based on viewer sentiment for recommendations.

🧪

Research to Production

From RL and optimization research (postdoc, PhD) to shipping real systems. I understand both sides of the research–engineering divide.

Generative AI Integration

Hands-on with OpenAI, Anthropic, open-weight LLMs, speech/video synthesis, and multimodal media models in production environments.

☁️

AWS ML Stack

Deep experience with SageMaker, Bedrock, AgentCore, Step Functions, Lambda, and CloudWatch AI Investigations.

Experience

Where I've worked

Aug 2023 – Feb 2026

Director, Research & Machine Learning

APMC
  • Built and led a cross-functional ML organization delivering production video intelligence systems across advertising, streaming, and generative media.
  • Owned ML strategy, architecture, and delivery across all initiatives; communicated progress, risks, and tradeoffs to all stakeholders
  • Implemented robust evaluation and safety frameworks, including RLHF, LoRa, and human-in-the-loop validation, to continuously measure and iterate on model accuracy, task completion, and to implement guardrails against content moderation and hallucination.
  • Architected and optimized Retrieval-Augmented Generation (RAG) pipelines and context graphs to ground agent responses in large-scale structured and unstructured data, ensuring factual accuracy and reliable information retrieval.
  • Designed and deployed agent orchestration architectures (using LangGraph/LangChain/AgentCore) for multi-step reasoning and tool execution, enabling autonomous decision-making and interaction with core business systems (e.g., infra/risk)
  • Designed and deployed neural network-based sequence modeling (including attention-based models) for temporal video analysis, bridging research and production to drive real-time metadata labeling and content intelligence
  • Implemented human-in-the-loop labeling workflows, closing the feedback loop between ML outputs and reviewer overrides to continuously improve model quality
  • Migrated team to AI-first Agile process, to have development lifecycle better leverage AI co-developers
  • Delivered organization-wide AI assistants enabling stakeholders to query KPIs and system state directly within existing workflows
  • Contributed as a developer and domain expert across all ML initiatives
AWS Sagemaker Bedrock AgentCore OpenAI Anthropic Multimodal LLMs Lambda
Jan 2023 – Aug 2023

Senior Researcher

APMC
Jan 2022 – Nov 2022

Senior Researcher

Oooh
  • Architected cross-modal integrations: text-to-speech, speech-to-text, text-to-image into video pipelines
  • Built content-based video recommendation and moderation systems
  • Developed video content scoring from human sentiment, for recommendations and segment highlighting
  • Improved low-latency live-streaming and P2P broadcast
  • Integrated and benchmarked client-side face/object detection, body tracking, and region-of-interest selection
Computer Vision Cross-modal AI Video ML Recommendation Systems
Sep 2018 – Jan 2022  ·  Tveon (3 yrs 5 mo)

Senior Researcher

Tveon
  • Led a team of video researchers building custom encoding tools and scalable compute infrastructure
  • Team spanned video engineers, developers, psychologists, neuroscientists, and data scientists
Python / OpenCV TensorFlow AWS Docker / Kubernetes
Nov 2012 – Sep 2018  ·  Tveon (5 yrs 11 mo)

Senior Cloud Architect

Tveon
  • Designed and led development of CMS and automated transcoding for live-streaming and critical infrastructure monitoring
  • Mobile app with location-aware auto-selection of live IP camera feeds for on-site security
  • Built custom encoding and CMS for multi-channel entertainment content with in-app micro-gaming
  • Established a talent pipeline from the local university's Math/CS program
Apr 2016 – Sep 2018

Director of Technology

Prosmart / Rosterbot / Sportgo
  • Architected the product as it expanded into four business verticals; integrated acquisition of Metabridge Top 15 startup Rosterbot
  • NLP-driven analysis of internal communications to measure community health
  • Co-developed a computer vision tool with UBCO partners to detect brand presence in user-uploaded images
  • ProSmart became the first company from the regional accelerator listed on the TSX
NLP Computer Vision AWS Rekognition
Jul 2018 – Oct 2018

Technical Consultant

Innovate BC
  • Consulted on the AI Justice Challenge with BC Ministry of Justice
  • Defined requirements and training data design for multi-modal AI applications (speech-to-text, translation, AI-guided retrieval)
Jan 2018 – Apr 2018

Lecturer — Digital Citizenship (COSC 101)

University of British Columbia
  • Taught 150–200 first-year students on digital platforms, online safety, cloud computing, and ML in social media
Jan 2011 – Dec 2021

Co-Organizer

Digital Okanagan / Startup Coffee / Data Science Meetup
  • Co-organized tech community events across Kelowna for 10+ years
  • Bridged UBCO faculty and local industry, with a current focus on AI advances
Jun 2011 – Oct 2012

Senior System Architect

Worldplay
  • Built a full-stack system to simulate and analyze 100k+ simultaneous video streaming sessions
  • Models for network issue simulation, user satisfaction tracking, and encoding quality feedback
Jan 2009 – Aug 2010

Postdoctoral Research Fellow

University of British Columbia
  • Research across Mathematics, Statistics, and Computing Science
  • Visualization of iterative projection method convergence; optimization on non-convex functions
  • Applied global optimization to open problems in statistical modelling
Optimization Non-convex Functions Statistical Modelling
2000 – 2002

Research Assistant

Packeteer Canada
  • Developed a (now-patented) method for re-encoding images to minimize visual artifacts relative to target filesize
  • Led human-subject experiments evaluating automated image encoding quality
Education

Background

PhD, Computer Science
Dalhousie University
2004 – 2008
PhD Candidate, Mathematics
Simon Fraser University
2002 – 2004
MSc, Mathematics
2000 – 2002
BA, Mathematics
University of British Columbia
1995 – 2000
Projects

Recent Work

A selection of projects across AI, video, and cloud infrastructure.

NCIPS
Non-Convex Iterated Projection Systems
Experimental interfaces for deriving equations for convergence conditions for non-convex IPSs.
Math Games
Math Games
A collection of interactive math visualizations and games. Inspired by my work in mathematical visualization and optimization.
Video Concepts
Video Concepts
Exploring the mathematical concepts underlying the intersection of video technology and creative expression.
Writing

Datamase

Observations on AI, cloud, and visual media — applications of AI to day-to-day life.

Cool Things in Tech, April 30-May 6, 2026
Cool Things in Tech, April 30–May 6, 2026
Overview report on the latest in AI and technology in the week of April 30–May 6, 2026, with a personal bias on generative media and society.
Cool Things in Tech, April 23-29, 2026
Cool Things in Tech, April 23–29, 2026
Overview report on the latest in AI and technology in the week of April 23–29, 2026, with a personal bias on generative media and society.
Visual Context Persistence in ChatGPT Images 2.0
Visual Context Persistence in ChatGPT Images 2.0
How visual context persistence in ChatGPT's image generation causes repeated visual elements to carry forward within a single session.
All posts on Substack → New posts on AI, video, and the tech world.
Contact

Let's talk

Whether it's an interesting problem, a new team, or just a good conversation about AI — I'm always happy to connect.