Mecka AI acquires Docula to process motion data

Companies training robots on human motion face two core bottlenecks: capturing diverse, real-world trajectories at scale and producing normalized, defensible datasets that downstream modelers trust. Acquiring small teams with proven high-volume data pipelines is a common shortcut to accelerate deployment of production-grade data infrastructure for physical AI.

Per reporting by BetaKit, Toronto- and New York City-based physical-AI startup Mecka AI acquired Vancouver startup Docula earlier this year. BetaKit reports that all three members of the bootstrapped Docula team joined Mecka and that deal terms were not disclosed. In an interview with BetaKit, Mecka special projects lead Mark Grinev said Docula's prior product focused on healthcare billing but handled "massive volumes of data."

BetaKit reports that Mecka collects three-dimensional physical motion data by distributing iPhones and custom cameras to hundreds of thousands of contributors across 12 countries, then processes and packages that data for customers training robotics systems. BetaKit also notes Mecka opened a New York office this month and that its headcount is 45, more than half of whom are Canadian. BetaKit cites a Robotics Center of Silicon Valley report that estimates the 2026 US robotics market at $11.4 billion USD, a near 30% year-over-year increase.

Companies assembling datasets for physical-AI commonly borrow tooling and personnel from adjacent, regulated domains where large-scale ingestion, normalization, auditing, and defensible reporting are core competencies. Docula's experience building an AI data-processing engine for medical billing, ingesting records, normalizing codes, running edits, benchmarking fees, and producing auditable reports, per BetaKit, maps to several of the operational problems robotics data teams face: schema harmonization, error-correction, and provenance tracking.

Watch for indicators that matter when evaluating third-party motion datasets or partnerships: dataset provenance and audit trails, normalization and label schema, temporal synchronization quality for multi-camera captures, contributor recruitment and diversity, and commercial licensing terms. Reporting by BetaKit provides the transaction and team details; Mecka has not disclosed financial terms.

This acquisition is a practical example of data-ops expertise being treated as strategic infrastructure for physical AI. For teams building or buying motion datasets, expect increased emphasis on robust ingestion pipelines and auditable normalization tooling drawn from regulated-data domains.