Home Technology About Team Research Blog Contact Request Evaluation Unit
About Tendonkindle

Built to close the actuator gap

We started in Pittsburgh in 2024 with a specific technical thesis: the joint actuator is the foundational bottleneck in humanoid robotics, and nobody had built the module that the field actually needed.

Founding Story

How we got here

James Whitaker spent three years as lead actuator engineer at a Pittsburgh robotics lab. During that time he watched two promising humanoid projects stall — not because of software limitations, not because of mechanical design failures, but because no commercially available joint module met their torque-to-weight requirements. Both projects spent six-month detours designing custom gearboxes before any meaningful locomotion work could begin.

The same pattern repeated across conversations with peers at other labs and early-stage robotics startups. Teams were independently rediscovering the same gap: industrial servo actuators were designed for fixed-arm assembly lines, not the compliance, backdrivability, and torque density that bipedal and quadrupedal platforms require. Every team hit the same hardware wall and then spent a year building around it.

In early 2024, Whitaker left to build the actuator module he wished had existed. The original insight was specific: the problem was not that the technology was impossible — strain-wave gearboxes, brushless motors, and MEMS force-torque sensors all existed. The problem was that nobody had integrated them into a sealed, production-oriented module with a co-designed software layer that made compliance control accessible from day one.

The first Tendonkindle prototype was assembled in Pittsburgh in early 2024 using a COTS brushless motor with a custom strain-wave reduction stage and an SPI-connected force-torque sensor. It proved out the torque-density target. The seal design and thermal management required a complete second-generation redesign, which took through late 2024. Two co-founders from Carnegie Mellon's Robotics Institute joined: Nadia Osei as CTO, bringing embedded control and real-time systems depth, and Rafael Tran as Head of Motion Systems, bringing experience from a prior industrial-robotics OEM.

Tendonkindle is now shipping Gen 1 evaluation units to four robotics labs for hardware-in-the-loop validation. The motion SDK beta is available to evaluation partners running ROS 2. No external funding has been raised as of mid-2026 — the company is bootstrapped by the founding team.

Mission

To eliminate the actuator bottleneck from humanoid robotics development — giving every team building bipedal and quadrupedal robots a production-grade, software-integrated joint module they can trust on day one.

The actuator gap is not a research problem. The physics are well understood — strain-wave gearboxes, brushless motors, and MEMS force-torque sensors all exist. The gap is a product problem: no one had integrated these components into a single sealed module, qualified it for the shock loads and thermal demands of legged locomotion, and shipped it with a software layer that makes compliance control immediately accessible to any team with a ROS 2 stack. That is precisely what Tendonkindle is building, one evaluation unit at a time.

Company Stage

Where we are

Pre-Seed — Hardware Validation

Tendonkindle is a pre-seed hardware startup currently in its hardware-in-the-loop validation phase. Gen 1 evaluation units are shipping to partner labs. The motion SDK beta is in active use. We are a founding team of three engineers, bootstrapped, with no external investors as of mid-2026.

The pre-seed stage means our priorities are concrete: validate the Gen 1 hardware spec against real locomotion loads, refine the SDK based on partner feedback, and prepare the Gen 2 design for broader distribution. We are not seeking press coverage or customer scale at this stage — we are engineering toward a hardware spec that teams can rely on.

Gen 1 evaluation units shipping to four partner labs
Motion SDK beta in active hardware-in-loop use with ROS 2
Founding team of three engineers, bootstrapped
Based in Pittsburgh, PA — Carnegie Mellon Robotics Institute proximity
Gen 2 module design in progress based on Gen 1 field data
Values

How we work

Three principles that govern our engineering decisions, our partner relationships, and our product specifications.

Precision over hype

Every specification we publish is measured from production hardware, not simulated. Torque density numbers, latency figures, and IP ratings all come from documented test protocols. We do not publish claims that cannot be independently verified by a partner with a torque bench and an oscilloscope.

Integration first

Hardware and software are co-designed from the same starting point. The impedance control layer is not a wrapper written after the motor driver was finished — it was specified as a requirement before the gearbox ratio was selected. This changes what the hardware can do at first boot versus after six months of custom firmware work.

Transparent specs

Evaluation partners receive full hardware datasheets, motor model files for MuJoCo simulation, and the test protocols used to derive every spec number. There are no hidden limitations, no undisclosed derating conditions, and no marketing-copy torque figures that apply only at zero speed and room temperature. Partners make decisions based on the same data our engineers use.

Get Involved

Work with us or apply for evaluation hardware

We are engaging with qualified robotics teams and research labs for Gen 1 evaluation. If your platform needs compliant, backdrivable joint actuators, let us talk.

Contact the Team