Facial Systems Consulting

Teams building facial AI often lack systematic ways to evaluate behavior. They can’t reliably tell which model is better, why expressions break, or where a problem originates in the pipeline. I design evaluation frameworks that make those decisions diagnosable, covering model state evaluation, data diagnostics, annotation standards, and the in-house documentation that keeps those systems usable after I’m gone.

The frameworks I bring to clients are drawn from building facial data infrastructure at scale. At Meta, I defined the annotation standards and collection protocols behind Quest Pro face tracking, securing three patents in avatar animation. Since then, I’ve developed model evaluation frameworks at Flawless AI and built expression detection standards at Affectiva.

AI / Tech

No two facial AI pipelines have the same problems. Before recommending anything, I start with a diagnostic, so I can understand your data, your models, your use cases, and where the gaps are. From there, engagements typically involve some combination of:

Model & data audits
Annotation strategy & documentation
Vendor evaluation & standardization
Gold set development
Automated performance measurement frameworks

Scope and depth depend entirely on where you are and what you’re trying to build.

Animation

Facial rigs are complex systems with many moving parts and many people interpreting them differently. When inconsistencies creep in across characters, departments, or pipelines, it’s rarely obvious where they started or how to fix them. I come in when teams need an outside eye with the anatomical and technical depth to diagnose where things break and define a path forward.

My rig and blendshape reviews are not simply a list of fixes. Each note is grounded in anatomy and muscle mechanics. Teams walk away understanding what to change as well as why, and how to apply that reasoning to every shape that comes after.

Some projects don’t have an existing playbook. When a studio needs a facial system for a capuchin, a chimpanzee, or a sports character with use-case-specific ROM requirements, the work starts with research. I investigate the anatomy, identify the gaps, and build a custom framework from scratch, iterating with modeling, rigging, and animation until the team can execute independently.