Banseog note · 2026-10-26
Turning hand movements on factory floors into robot training data
Current stage · from public events · 2026-11-14
- R&D · prototype
- Pilot · first production
- Scaling production
- Operations · field rollout
- Going global · localizing
- M&A · integration
- Restructuring
Operations · field rollout Collecting and processing real-world motion data for robotics company customers
- 2026-10-17 In an interview, the company described securing its first customers and serving robotics companies in the YC portfolio, and explained how data collected and processed in Uzbekistan is supplied to the US market. Source
Roles often seen at this stage Robot software, AI & vision · Field, integration & service · Quality, regulatory & clinical
Companies at the same stage → · Same stage in other industries →
- With headquarters in San Francisco and a second office in Tashkent, Uzbekistan, it says 20 capture rigs are recording at sites in eight industries, including manufacturing, warehousing, kitchens, hotels, construction and mining, and that it is deploying 50 more.
- Its public dataset EgoViz-120 covers 120 hours and 5,392 clips and is released under CC-BY-4.0.
- In an October 2026 interview, founder Bobirjon Mardonov said the company met its first customer at NVIDIA GTC, that its customers include robot companies in the Y Combinator portfolio, and that the structure has data collection handled in Uzbekistan and the main market in the US.
Decisions within a year
- AKFA Group pilot the company says it has started an exclusive pilot program with the AKFA Group, which it describes as one of the largest companies in Central Asia.
- More capture rigs it says it is deploying 50 more on top of the 20 currently recording.
- Tashkent hiring in September 2026 it posted full-time Tashkent-based roles for a robotics engineer (designing and maintaining capture rig and sensor hardware) and a data scientist.
People they will look for
- Capture rig and sensor hardware engineer — someone to handle RGB-D camera calibration, 3D-printed prototypes, wiring and assembly as 50 more rigs are deployed
- Robot learning data scientist annotation and quality pipelines that turn hand keypoints, depth and IMU signals into action steps and natural-language descriptions
- Field data collection operations manager — planning worker recording sessions at sites in eight industries and keeping quality standards
For hiring managers
In Korea, humanoid developers, robot research units at large companies and robot learning data startups look for the same sensor and data talent. At a stage when capture rigs are being greatly expanded, it is safer to secure people who understand rig calibration and data quality before the rigs come in.
For job seekers
Experience with RGB-D camera calibration, multi-sensor synchronization and data formats for robot learning (LeRobot, RLDS, HDF5) is worth the most right now. The public openings are full-time roles based in Tashkent, so please check the work location conditions first.
An analysis written from public sources. Sources: Humaid website (checked October 2026; offices, 20/50 capture rigs, AKFA pilot, EgoViz-120), Hugging Face humaidtech/EgoViz-120 (updated August 31, 2026), Humaid careers page (September 2026, Tashkent robotics engineer), Daryo 2026-10-17 (founder interview; first customer, customer base, collection structure).
Roles hiring now
Skills asked for most
Open jobs
Banseog checks official company careers pages every day. A job is marked closed only when the official page returns a removal (404 or 410).