# Bucket Robotics Expands CAD-Driven Vision From Auto Plants to Air Force Depots

**Author:** Zero G Talent
**Published:** 2026-11-19
**Category:** engineering
**Tags:** yc-hiring, computer vision, manufacturing, synthetic data, robotics, software engineering

[Read on Zero G Talent](https://zerogtalent.com/blog/bucket-robotics-expands-cad-driven-vision-from-auto-plants-to-air-force-depots)

---

Automating surface inspection in high-mix manufacturing has traditionally stumbled over a single bottleneck: training computer vision models requires thousands of real photos of defective parts that factories rarely log. [Bucket Robotics](https://zerogtalent.com/robotics-companies/bucket-robotics) bypasses physical defect collection entirely by generating photorealistic synthetic datasets directly from 3D CAD files. 

After demonstrating its surface inspection setup at the 2026 Consumer Electronics Show, as reported by [TechCrunch](https://techcrunch.com/2026/01/18/how-yc-backed-bucket-robotics-survived-its-first-ces/), the seed-stage company is actively scaling from commercial automotive components into military aviation sustainment. Backed by Y Combinator (S24), Soma Capital, Pioneer Fund, Lightscape Partners, and Taro Fukuyama, the Pittsburgh-headquartered startup operates with four employees and is now expanding its engineering roster in San Francisco.

## Generating Defect Data Without Physical Scraps

Bucket Robotics focuses on defects that manual inspectors typically spot on production lines: surface scuffs, burns, warpage, and structural tears. In an interview with [TechCrunch](https://techcrunch.com/2026/01/18/how-yc-backed-bucket-robotics-survived-its-first-ces/), CEO [Matt Puchalski](https://zerogtalent.com/people/matt-puchalski) noted that for parts like automotive door handles, interior structural inspection is largely solved, but flawless exterior cosmetic inspection remains difficult to automate without massive training sets. Auto manufacturers, Puchalski said, usually end up deploying human workers to manually inspect surfaces.

To remove that cold-start hurdle, Bucket Robotics ingests standard STEP and CAD files into an internal pipeline called WrenchML. As demonstrated in a [Bucket Robotics product video](https://www.youtube.com/watch?v=tr9LvR0nF2A), users upload CAD models into a browser-based interface, select material finishes such as molded plastics, and configure simulated defect distributions, including surface bumps, burns, or warping. The system renders between 5,000 and 25,000 synthetic images showing the part under varying lighting conditions, backgrounds, and camera sensor parameters.

Once synthetic models are trained, Bucket exports them to ONNX formats for edge inference. In a [defect detection deep dive](https://www.youtube.com/watch?v=RCyguguf3Is), co-founder and CTO [Stephan Wolski](https://zerogtalent.com/people/stephan-wolski) showed the inspection stack running real-time inference on a $400 NVIDIA Jetson board paired with a $50 camera module, verifying production parts without proprietary sensing rigs. The stack also interfaces directly with factory Enterprise Resource Planning (ERP) and Manufacturing Execution Systems (MES) to automate disposition records.

## Autonomous Vehicle Pedigree Applied to Dual-Use Inspection

Both founders bring backgrounds from high-stakes autonomous vehicle programs. Before co-founding Bucket Robotics, [Matt Puchalski](https://zerogtalent.com/people/matt-puchalski) worked on AV integration and operations across Uber ATG, Argo AI, Latitude AI, and SoftBank-backed Stack AV. Co-founder and CTO [Stephan Wolski](https://zerogtalent.com/people/stephan-wolski) started college at age 13, served as Remote Guidance Tech Lead at Argo AI, and built foundational autonomy infrastructure at Stack AV.

Their experience operationalizing autonomy in complex environments has pushed Bucket Robotics toward dual-use contracts. Outside automotive and commercial plastics plants, the startup has deployed its machine learning software, dubbed InspectHub, into defense maintenance environments.

In a submission video for [Tradewinds AI](https://www.youtube.com/watch?v=ojyVU5mHPfo), the company demonstrated InspectHub operating at Technology Readiness Level 6 (TRL-6) at the Warner Robins Air Logistics Complex. Operating under a Phase II Small Business Innovation Research (SBIR) transition and an STTR award, InspectHub analyzes non-destructive inspection (NDI) images—including x-ray, ultrasonic, and thermal scans—for aircraft like the A-10 and F-15. By running sensor-agnostic AI pre-screening on premises to identify potential foreign object debris (FOD) indications, the software aims to replace manual paperwork and CD-ROM transfers while reducing defect tagging workflows by more than 50%.

## Engineering Roles Across the Pipeline

Bucket Robotics lists three open technical positions, all based onsite in San Francisco, CA, US. The roles reflect the company's division between edge integration on shop floors, synthetic vision modeling, and enterprise software surfaces.

Deploying on real factory floors requires hands-on systems integration. The [Forward Deploy Engineer](https://zerogtalent.com/robotics-jobs/bucket-robotics/forward-deploy-engineer-95797) ($100,000—$120,000) integrates cameras, machine learning models, and workflows into live manufacturing lines to debug and harden systems under factory conditions, requiring up to 30% travel to customer facilities across the United States.

On the model development side, the [ML Engineer](https://zerogtalent.com/robotics-jobs/bucket-robotics/ml-engineer-73634) ($120,000—$165,000) designs and trains CAD-native computer vision models that convert 3D geometry and synthetic renderings into production inspection tools, optimizing them for low-latency edge deployment.

To support internal infrastructure and external customer platforms, the [Senior Full Stack Engineer](https://zerogtalent.com/robotics-jobs/bucket-robotics/senior-full-stack-engineer-95125) ($160,000—$220,000) builds the customer-facing interface, internal deployment utilities, and data pipelines connecting CAD ingestion with computer vision systems.

## Open Positions at Bucket Robotics

| Role | Location | Work Mode | Salary |
| --- | --- | --- | --- |
| [Forward Deploy Engineer](https://zerogtalent.com/robotics-jobs/bucket-robotics/forward-deploy-engineer-95797) | San Francisco, CA, US | Onsite | $100,000—$120,000 |
| [Senior Full Stack Engineer](https://zerogtalent.com/robotics-jobs/bucket-robotics/senior-full-stack-engineer-95125) | San Francisco, CA, US | Onsite | $160,000—$220,000 |
| [ML Engineer](https://zerogtalent.com/robotics-jobs/bucket-robotics/ml-engineer-73634) | San Francisco, CA, US | Onsite | $120,000—$165,000 |

[See all open roles at Bucket Robotics](https://zerogtalent.com/robotics-companies/bucket-robotics)

---
*Source: [Zero G Talent](https://zerogtalent.com), the job board for space, defense, AI, robotics and frontier-tech companies.*