Robots to Automate Fulfillment

Yondu is creating the robotic workforce of the future starting with logistics automation. We're deploying humanoid robots in the first flexible, drop-in picking automation solution designed for 3PLs.

Active Founders

Michael Chen

Founder

Michael Chen is at a unique crossroads with a background in intelligent machines from MIT and research experience at the forefront of robotics and AI. He helped build an open source bipedal robot at MIT CSAIL. Michael is an avid entrepreneur who has always loved business (even owning a sauce brand). For fun, he builds all sorts of cool machines ranging from FPV drones to train track cleaning robots to gasoline powered snowboards.

Michael Chen

Founder

Michael Chen is at a unique crossroads with a background in intelligent machines from MIT and research experience at the forefront of robotics and AI. He helped build an open source bipedal robot at MIT CSAIL. Michael is an avid entrepreneur who has always loved business (even owning a sauce brand). For fun, he builds all sorts of cool machines ranging from FPV drones to train track cleaning robots to gasoline powered snowboards.

Tahmid Jamal

Founder

Tahmid graduated from MIT with degrees in AI and AeroAstro. There he's worked on projects ranging from language model compression and quantization to astrodynamics simulations for a solar research proposal to NASA. He was doing his master's researching uncertainty estimation methods in deep learning before dropping out to participate in YC. In his free time, he enjoys movies, board games, and climbing the occasional tree.

Tahmid Jamal

Founder

Tahmid graduated from MIT with degrees in AI and AeroAstro. There he's worked on projects ranging from language model compression and quantization to astrodynamics simulations for a solar research proposal to NASA. He was doing his master's researching uncertainty estimation methods in deep learning before dropping out to participate in YC. In his free time, he enjoys movies, board games, and climbing the occasional tree.

Company Launches

Yondu AI ➡️ Enabling the robot workforce of the future

See original launch post

We are a platform-to-application company that puts robots in warehouses and makes them useful on day one. Our state-of-the-art remote teleoperation gives immediate efficiency gains, with our robot behavior models gradually taking over to reach full automation and drive costs down further. We’re starting with mobile order fulfillment, where a huge chunk of warehouse labor spend sits.

uploaded image

https://www.youtube.com/watch?v=GzglTuRloCo

——————————————————————————————

The timing

General purpose robot hardware has gotten 10x cheaper in the past decade. And advancements in robot learning are helping automate long-form, complex tasks. BUT it’s still hard to achieve the perfect human level success rates necessary for deployment! The opportunity lies in automating the shorter-form, easier-to-master tasks that still have a level of complexity that made them impossible to automate in the past. The logistics field is full of these types of problems…

uploaded image

A simple, fully autonomous pick at 1× speed

——————————————————————————————

The problem in mind

Order picking labor can account for up to 50% of warehouse manual operating costs. Pickers often walk miles per shift! Warehouses still rely on people for repetitive, physically demanding tasks with high turnover. Existing order fulfillment automation falls into two buckets:

- ASRS (Automated Storage and Retrieval Systems): Basically turns a warehouse into a giant 3D robot grid system. It gets high throughput but costs tens of millions and requires an overhaul of warehouse space. 😢 HIGH COST, 😊 HIGH AUTOMATION

- AMR (Autonomous Mobile Robots): These robots carry shelves and goods around which reduces walking time. But they still require human picking. 😊 LOW COST, 😢 LOW AUTOMATION

——————————————————————————————

OUR SOLUTION

We’ve created a 3 part embodied-AI platform that automates bin picking using off the shelf robots in brownfield deployments. 😊 LOW COST, 😊 HIGH AUTOMATION

https://youtu.be/UVmMSui2_FQ

With ULTRON (user-friendly low-latency remote teleoperation system) we deploy first, putting robots to use immediately in our customers’ operations and spinning up our data flywheel.

uploaded image

Our WMS (warehouse management software) integration and orchestration software routes orders from a warehouse’s WMS to our platform, instructing robots where to go and operators what to pick.

uploaded image

Our autonomy stack YGM (Yondu General Manipulation) pushes the system to fully autonomous as we gain more in-the-wild experience. Over time the manipulation policy takes over more work and the operator is required less. YGM is responsible for:

Steering our robot policy to pick and place from the correct bins
Identifying how good a job it is doing and whether it is struggling so an operator can take over
Learning from new experiences and adapting to succeed in situations that it previously struggled in

uploaded image

Operations in our partner warehouse

——————————————————————————————

Our Vision

Our platform is not limited to only helping us conquer bin picking. Logistics warehouses still have many more repetitive-yet-slightly-complex tasks such as decanting, replenishing, and packing. And as robot learning capabilities continue to grow, our pipeline of deploy→collect→refine→automate only increases in value. There will be more brownfield robotics automation providers for specific industries and increasingly complex tasks. They will need to integrate with their customers and spin up their own data flywheels. And the Yondu platform will be available for them so they can automate any task with any robot.