Humynex builds the first cadaver-validated, think-aloud-annotated cognitive-motor training dataset for soft-tissue surgical robotics AI — starting with liposuction, from a surgeon who has performed it ten thousand times.
The surgical robotics market is scaling rapidly. The hardware exists. The AI frameworks exist. The missing layer is training data that captures why an expert surgeon makes each decision — not just what the instrument did.
Hospital recordings are fragmented, de-identification-constrained, and almost never annotated with expert decision logic. What exists is kinematic logs — instrument position and force — without the cognitive layer that explains each micro-decision.
Soft tissue deformation during liposuction — the interaction between cannula pressure, fat planes, fibrous septa, and fluid dynamics — is one of the hardest physical simulation problems in surgical robotics. Domain randomization approaches that work for rigid manipulation break down here.
The operative decision logic of a surgeon with 10,000 repetitions is tacit knowledge. It has never been systematically time-aligned to instrument signals and formatted as supervised learning data. That externalization is what Humynex provides.
"The slow progress toward surgical robot autonomy can be attributed to a few key issues: the scarcity of large, open-source datasets for training, challenges in modeling soft-body deformations encountered during surgeries, and the increased risk of patient injury during clinical trials."
Yip et al. — Science Robotics, 2024
A Humynex cognitive-motor surgical record is a time-indexed bundle of five signal streams captured during expert cadaver demonstration sessions — conducted under willed body program tissue use agreements. No IRB required. No patient consent burden. First session executable within 30 days of facility access.
Egocentric + field view of the operative site and instrument at 30fps
Inline cannula pressure telemetry — the tactile signal driving tissue-plane decisions
3-axis accelerometer & gyroscope on the cannula handle; trajectory and velocity
Expert surgeon describing decisions in real time — "reducing suction; deep fascia resistance"
Discrete skill-boundary labels: tissue_plane_transition, cannula_swap, abort_pass
Dr. Matlock is the primary data asset of Humynex. A board-certified physician and entrepreneur with an MBA from UC Irvine Paul Merage, he has performed over 10,000 liposuction procedures across four decades of practice and trained 435 surgeons in 46 countries in his pioneered techniques.
During Humynex cadaver capture sessions, Dr. Matlock performs expert demonstrations while narrating every operative decision in real time — tissue plane identification, suction pressure adjustment, cannula selection, abort criteria. That narration, synchronized to force and motion telemetry, is the supervisory signal that current surgical robotics AI programs cannot generate from simulation or clinical video.
He is also the President & CEO of Beverly Hills Sunset Surgery Center (1994–Present) and the Laser Vaginal Rejuvenation Institute of America (2003–Present) — two medical businesses he built and continues to operate.
A machine learning scientist serves as technical advisor, providing expertise in multimodal dataset design, synchronization pipelines, and surgical robotics AI model architectures — including imitation learning, diffusion policy, and vision-language-action model fine-tuning. The advisor's near-term role is to define the minimum viable capture schema, write the processing pipeline, and co-author the first technical publication. Advisor identity available under NDA on request.
The Humynex capture platform converts simultaneous sensor streams into structured, machine-learning-ready records through four sequential stages.
10,000-case surgeon performs the procedure with real-time think-aloud narration. Willed body program tissue use agreement. No IRB pathway required.
RGB-D camera, IMU on cannula handle, inline pressure transducer, directional microphone. Sub-$3,000 MVP hardware stack. Setup under 30 minutes.
Hardware-level timestamp alignment across all five modalities. Automated quality checks flag drift events before data enters the training pipeline.
Time-indexed records: video frame + force + motion + narration + event tag, all aligned to the same timestamp. Direct ingestion by standard ML frameworks.
| Data Source | Expert Level | Intent / Decision Layer | Tissue Validity | Humynex Advantage |
|---|---|---|---|---|
| Public OR video recordings | Variable — includes trainees | ✕ None | ✕ Live patient; consent-limited | ✓ on all three |
| Dry-lab phantom studies | Mixed | ✕ None | ✕ Synthetic tissue; poor fidelity | ✓ on all three |
| da Vinci kinematic logs | ~ Surgeon-level | ✕ No narration or decision tags | ✕ Clinical; no cadaver control | ✓ adds cognitive layer |
| Simulation only | N/A | ✕ None | ✕ Sim-to-real gap unvalidated | ✓ real tissue + real expert |
| Humynex Dataset | ✓ 10,000-case expert | ✓ Full think-aloud + event tags | ✓ Willed body cadaver — anatomically valid | Unique in market |
The AI in robot-assisted surgery segment is growing at 44% CAGR. Intuitive Surgical's proprietary dataset is platform-locked and not commercially available. No independent company has published an expert-annotated, cadaver-validated surgical cognitive-motor dataset.
From $5.5B in 2024, growing at 44.3% CAGR — driven by AI-guided autonomy, intraoperative decision support, and training data platforms. Source: Fortune Business Insights, 2024.
Intuitive Surgical performed 2.68M procedures in 2024 — 17M total. 2,000+ U.S. hospitals operate robotic surgical systems. The hardware infrastructure is already deployed at scale.
The data infrastructure layer is unoccupied. No independent company has built an open, expert-annotated, cadaver-validated surgical cognitive-motor dataset. That is the gap Humynex fills.
Intuitive Surgical, Medtronic, Stryker, J&J (Ottava) — all developing next-generation AI-guided platforms that require expert decision-layer training data.
Stanford CHARM, JHU LCSR, CMU Biorobotics — established surgical robotics labs with active need for annotated training data and limited access to expert operators.
NIH SBIR Phase I (~$300K), DARPA RSTAS (Robotic Surgical Training and Assessment) — directly aligned program areas with active funding cycles.
A proprietary expert surgical dataset becomes a strategic acquisition target as OEMs compete on AI differentiation. Intuitive's da Vinci 5 force feedback introduction signals this competitive vector.
Humynex is raising $500K pre-seed. SAFE note, $4M cap, 20% discount. Target close Q2 2026. Seed trigger: published dataset paper + one OEM pilot LOI.