Tangent Robotics Raises $4.5M for Robot Dexterity

The last few millimeters of an industrial assembly can keep a person inside an otherwise automated production line. Tangent Robotics has raised a $4.5M pre-seed round to move dexterous robot-hand research toward that demanding manufacturing work, with Fly Ventures and Toyota Ventures co-leading the September 30, 2026 financing and Logos Fund and Sparked Ventures participating.

The New York company is building what it calls a fine-motor-skill layer for robots: a combination of custom hand hardware, tactile sensing, teleoperation, data collection, and motor learning. The opportunity sits inside the final contact-rich steps of production, where a machine must thread, seat, snap, mate, wire, inspect, or reassemble a component without turning a small alignment problem into a quality failure.

That makes the round about more than adding another hand to the physical-AI market. Tangent is financing the transition from a strong Columbia Engineering research lineage and public prototypes toward repeatable skills that have to survive factory cycle times, maintenance requirements, changing parts, and imperfect real-world conditions.

What Tangent Robotics Raised

Tangent Robotics announced the $4.5M pre-seed financing as its initial funding round. The company said the capital will support development and deployment of skilled robotic manipulation at the intersection of hand hardware, touch sensing, and motor learning in demanding manufacturing environments.

Fly Ventures and Toyota Ventures co-led the round. Logos Fund and Sparked Ventures also participated, while Toyota Ventures partner David Sokolic led his firm's investment. Neither Tangent nor its investors disclosed a valuation, customer roster, revenue figure, production deployment count, or detailed allocation of the proceeds.

The company was founded in 2022 as a spinout from Columbia Engineering's Robotic Manipulation and Mobility Lab. Its current leadership includes Pedro Piacenza, co-founder and CEO; Matei Ciocarlie, co-founder and Chief Science Officer; Ioannis Kymissis, co-founder and consultant for unconventional electronics; and Tianjian Chen, CTO.

Why Dexterity Still Blocks Automation

Industrial automation is very good at motion that can be tightly specified and repeated. It becomes less reliable when the job depends on contact, subtle force, changing geometry, and recovery from a nearly correct position. A part can arrive at the right station and still need a person to feel whether the connector seated, the gasket folded, the thread caught, or the wire found the correct path.

Tangent's answer is not to reproduce the human hand in metal. The company is designing non-human articulation around the task, then combining that hardware with light-based touch sensors and learning systems. High-level vision-language-action models may determine what should happen, while Tangent's proposed layer handles the physical exchange between intention and the material world.

The company identifies assembly, threading, connector mating, gasket seating, snap fitting, gear meshing, wiring, installation, inspection, maintenance, and reassembly as target tasks. Those are narrow movements, but they carry broad economic consequences because one stubborn manual step can limit how much of a production process can actually be automated.

From Columbia Research to a Factory Product

Tangent's technical history is unusually visible for a pre-seed company. In 2020, Columbia Engineering documented an optical tactile finger that localized touch at sub-millimeter precision across a multicurved surface. The design measured how light moved through a transparent layer when the finger's skin deformed, then used data-driven methods to infer contact position and force.

Columbia later demonstrated a dexterous research hand operating without external cameras. The hand used tactile and proprioceptive data with motor-learning methods to manipulate an object in hand, and TIME included the Columbia University robotic hand among its Best Inventions of 2023.

The founders' work also extends beyond the current prototypes. Columbia describes hand and gripper mechanisms developed for NASA's Astrobee robots aboard the International Space Station, while Ciocarlie's earlier work at Willow Garage included contributions to the Robot Operating System. That history supports the team's credibility, but research demonstrations are not the same thing as a maintained commercial system inside a live production environment.

What the Investors Are Backing

Toyota Ventures describes Tangent as a way to address the contact-rich last mile of manufacturing. Its investment thesis emphasizes a pragmatic hardware approach in which sensing carries more of the burden instead of forcing every capability into a mechanically complex imitation of human anatomy.

Fly Ventures lists Tangent in industrial automation and focuses on technically difficult companies at pre-seed and Seed. The co-leads are backing the possibility that physical AI needs an intermediate layer between semantic reasoning and low-level control, especially when a robot must learn a manipulation skill and repeat it under production constraints.

Tangent says its goal is to reduce the practice time for complex motor skills to days and eventually hours. That target is commercially important because manufacturers cannot wait through human-scale learning curves every time a task, component, or line changes, but Tangent has not published audited results showing that the target has been reached.

What the $4.5M Still Has to Prove

The company's next evidence will have to come from manufacturing, not another clean prototype video. Buyers will care about repeatability, cycle time, force control, recovery from errors, hardware life, integration cost, maintenance, operator training, and the percentage of task variation the system can handle without an engineer rebuilding the policy.

Tangent has not named customers or disclosed commercial deployments, revenue, valuation, or independently audited task metrics. That absence is normal at pre-seed, but it defines the work the financing must now support. The technical question is whether touch, articulation, and learning can become a system that keeps working after the demonstration ends and the production line keeps moving.

The company's strongest argument is that physical intelligence is decided at contact. A robot may understand the sequence, locate the part, and reach the station, yet the economic value appears only when it can complete the small handoff without damaging the component, stopping the line, or summoning a person to finish the job. Tangent's $4.5M gives the team a chance to make those last few millimeters dependable enough to belong to the machine.