The most valuable knowledge on a factory floor is often the hardest to capture. It is the tacit, physical expertise of a veteran operator, the sequence of motions and decisions that keeps a line running. DeepHow, a Detroit-based AI company founded in 2018, is betting that video, not a written manual, is the key to bottling that knowledge. Its platform uses AI to transcribe, segment, and index videos of skilled work, turning them into searchable, step-by-step instructions and training modules for frontline workers [PERPLEXITY SONAR PRO BRIEF].
For multinational manufacturers facing a skilled labor shortage, the pitch is direct. The company claims its platform can reduce the time to create training content by 10x and improve worker performance by 25% [Purpose Jobs, September 2021]. By April 2023, DeepHow reported over 70 global customers, a majority of them multinational enterprises [PR Newswire, April 2023]. The company calls its approach "Physical AI," a term that underscores its focus on understanding and verifying actions in the real world, not just processing text [deephow.com].
The Wedge: From Video to Verified Workflow
DeepHow's core product is an AI "workflow indexer." A worker records a task on a phone or tablet. The AI then transcribes the audio, identifies key steps, annotates them, and organizes the workflow into a reusable digital asset [PERPLEXITY SONAR PRO BRIEF]. This creates what the company markets as "Smart Videos" and "Smart Instructions."
The evolution beyond simple documentation is where the enterprise value proposition sharpens. The platform's generative AI features can assist in creating standard operating procedure (SOP) videos and provide chat-based assistance [PR Newswire, April 2023]. More critically, its "Time and Motion AI" and "Live SOP Verification" aim to move from recording work to analyzing and ensuring it happens correctly. This includes identifying production bottlenecks, verifying task execution in real time, and automating traditional time-and-motion studies [deephow.com].
A Team Forged in Industrial R&D
The founding team's background reads like a blueprint for this specific problem. CEO Samuel Zheng and CTO Wei-Liang Kao are both PhDs who worked together at Siemens Corporate Research [Sierra Ventures, September 2021]. Zheng's doctorate is in Human Factors and Engineering Psychology, while Kao's is in Computer Information and Technology. They co-founded the company with Patrik Matos da Silva in 2018 after identifying the acute need for knowledge transfer in industrial settings [PERPLEXITY SONAR PRO BRIEF].
This Siemens pedigree provided more than just domain insight. The team was selected for the Techstars Mobility accelerator program and has since built a roster that includes engineering, product, and customer-facing roles, suggesting a shift from pure R&D to commercial execution [LinkedIn, retrieved 2026].
Traction and the Regulated Frontier
DeepHow's customer base and partnerships signal adoption beyond early pilots. The company does not disclose revenue, but its disclosed funding and customer count outline a venture-scale trajectory. The traction is underscored by a recent, specific partnership.
In July 2026, Yazaki North America, a major automotive components supplier, announced an agreement to deploy DeepHow's Time and Motion AI across its production lines in Michigan [MarkLines Automotive Industry Portal, July 2026]. The deployment will utilize NVIDIA technologies, pointing to the computational demands of the vision-based AI [globalautoinsight.com].
Perhaps more telling for long-term durability is DeepHow's push into regulated manufacturing. The company offers "PharmaCloud," a dedicated, GMP-compliant SaaS environment for pharmaceutical and medical device manufacturers [DeepHow Blog]. It has also partnered with the University at Albany’s College of Nanotechnology, Science, and Engineering to adapt its platform for semiconductor manufacturing training [University at Albany, 2024]. These moves target sectors where documentation, compliance, and training rigor are non-negotiable,and costly.
Pre-Series A (Sep 2021) | 9 | M USD
Series A (Apr 2023) | 14 | M USD
The Capital Stack and Competitive Landscape
DeepHow has raised a total of $23 million across two primary rounds. Sierra Ventures led a $9 million pre-Series A in September 2021 [DBusiness, September 2021]. In April 2023, the company closed a $14 million Series A with a syndicate that included Foothill Ventures, LG Technology Ventures, Osage Venture Partners, Owl Ventures, and Qualcomm Ventures, alongside returning investor Sierra Ventures [PR Newswire, April 2023].
The competitive field for industrial training and workflow software is crowded, but DeepHow's video-first, AI-native approach carves a distinct niche. The company's public materials do not name direct competitors, focusing instead on displacing manual methods, outdated video libraries, and static PDF work instructions. The risks are inherent in any deep-tech SaaS play targeting large, slow-moving industrial customers.
- Implementation complexity. Deploying vision AI and workflow analysis on a factory floor is harder than shipping a web app. It requires integration with existing manufacturing execution systems (MES) and learning management systems (LMS), a challenge DeepHow says it addresses with pre-built connectors [Techstars Job Board].
- Proof at scale. While the Yazaki deal is a significant proof point, the broader claim of "thousands of global users" from 2021 needs contemporary validation against the 70+ enterprise customer count [Sierra Ventures, September 2021]. Enterprise sales cycles are long, and land-and-expand motion within each account will be critical for unit economics.
- Defensibility. The core AI technology for video segmentation and analysis is advancing rapidly. DeepHow's moat may lie less in a proprietary model and more in the industry-specific workflows it encodes, the regulatory compliance frameworks like PharmaCloud, and the accumulated library of indexed industrial knowledge.
The company's answer to these challenges appears to be specialization and depth. By focusing on physical workflows and regulated verticals, DeepHow is building a domain-specific data asset that a general-purpose video tool cannot easily replicate.
What the Next Twelve Months Will Reveal
The $14 million Series A closed nearly three years ago. For a company investing heavily in R&D and enterprise sales, that runway dictates a near-term milestone. The logical next step is a Series B round to accelerate growth, likely predicated on demonstrating expanded deployments within its existing multinational roster and landing more flagship partnerships like Yazaki.
The key metric to watch will be customer concentration and expansion. Do the 70+ enterprises represent 70 single-site pilots, or have they grown into multi-plant, multi-use case deployments? The partnership with the University at Albany suggests a focus on the high-stakes semiconductor sector, another potential growth vector.
DeepHow's bet is that the factory floor's most valuable asset,human expertise,is also its most fragile. With backing from Sierra Ventures, Qualcomm Ventures, and LG's venture arm, and a product now in use at names like Foxconn and ISAIC according to vendor reports, the company has convinced investors its AI can make that knowledge permanent, searchable, and actionable [Airframe]. The question for the next check-writer is whether that captured expertise can scale as predictably as the assembly lines it aims to optimize.
Sources
- [PR Newswire, April 2023] DeepHow Raises $14M Series A Funding to Transform How Enterprises Capture and Transfer Technical Know-How | https://www.prnewswire.com/news-releases/deephow-raises-14m-series-a-funding-to-transform-how-enterprises-capture-and-transfer-technical-know-how-301809091.html
- [Sierra Ventures, September 2021] DeepHow - Why Sierra Ventures Invested | https://www.sierraventures.com/content/deephow-why-sierra-ventures-invested
- [Purpose Jobs, September 2021] Detroit startup DeepHow raises $9M pre-Series A funding | https://www.purpose.jobs/blog/detroit-startup-deephow-raises-9m
- [DBusiness, September 2021] DeepHow in Detroit Raises $9M Pre-Series A Funding to Bridge the Skills Gap | https://www.dbusiness.com/tech-mobility-news/deephow-in-detroit-raises-9m-pre-series-a-funding-to-bridge-the-skills-gap/
- [deephow.com] Physical AI for Operational Excellence | AI Knowledge & Verification | https://deephow.com/
- [PERPLEXITY SONAR PRO BRIEF] DeepHow company briefing
- [MarkLines Automotive Industry Portal, July 2026] DeepHow and Yazaki North America agreement announcement | https://www.marklines.com
- [globalautoinsight.com] Yazaki North America to deploy DeepHow Time and Motion AI | https://globalautoinsight.com
- [University at Albany, 2024] University at Albany partners with DeepHow for semiconductor training | https://www.albany.edu
- [Airframe] Vendor-reported customer deployments for Foxconn and ISAIC
- [Techstars Job Board] DeepHow company profile | https://jobs.techstars.com/companies/deephow
- [LinkedIn, retrieved 2026] DeepHow team profiles
- [DeepHow Blog] PharmaCloud product description | https://blog.deephow.com/