Ashish Kumar, the former lead of Tesla's Optimus AI team, has co-founded a new robotics company called Intelligent Machines, based in San Francisco. The startup is building non-humanoid robots meant to perform tasks normally handled by highly skilled workers in manufacturing and aerospace, according to a report from The Information via Dealroom. The company has already pulled in several engineers who previously worked with Kumar on Optimus, Tesla's humanoid robot program, and is in the process of raising a seed round that could reach roughly $100 million — an unusually large opening check for a company at this stage.
Table of Contents
- Who Is Behind Intelligent Machines?
- What Makes These Robots "Non-Humanoid"?
- How the Robots Learn Their Skills
- Why a $100 Million Seed Round Stands Out
- What This Means for the Industrial Robotics Field
- Frequently Asked Questions
Who Is Behind Intelligent Machines?
Ashish Kumar spent a little over two years at Tesla, roughly a year of it as the AI lead on the Optimus humanoid robot team, before announcing on September 19, 2025 that he was leaving the company. At the time, he said his next step was joining Meta as a research scientist, a move confirmed in Sawyer Merritt's September 2025 post and covered by Benzinga. That stint at Meta turned out to be brief.
Reporting on the new startup indicates Kumar spent about four months there before leaving to start Intelligent Machines, according to The Information's reporting via Dealroom. His technical approach traces back further than either job: his PhD research at UC Berkeley focused on reinforcement learning, the same method Intelligent Machines is now applying to teach robots complex tasks. Several of Kumar's former Optimus colleagues have already joined him at the new company, giving Intelligent Machines a team with direct experience building humanoid robot software at Tesla's scale, even though the new venture is explicitly moving away from the humanoid form factor.
What Makes These Robots "Non-Humanoid"?
Optimus, the project Kumar led at Tesla, is a humanoid robot — built with a human-like torso, arms, and legs so it can work in spaces and with tools designed for people. Intelligent Machines is taking a different path, building robots that are not shaped like humans but are instead designed around the specific tasks they need to perform.
The focus is on jobs currently done by highly skilled workers in manufacturing and aerospace, fields where precision and specialized technique matter more than general-purpose mobility. By skipping the humanoid form, the company can tailor a robot's physical design to a narrower set of tasks, rather than building a general-purpose body first and teaching it skills second.
How the Robots Learn Their Skills
Intelligent Machines is currently focused on software rather than hardware, using reinforcement learning — a training method where a system improves by trial and error, guided by feedback on whether its actions succeed — to teach robots complex, skilled tasks. This mirrors the subject of Kumar's doctoral research at Berkeley.
Company leaders have discussed designing their own hardware eventually, but for now the software-first approach lets the team develop and refine the underlying skills before committing to a specific robot body, according to the Dealroom report. That sequencing — training the intelligence first, building the machine around it later — is a departure from how Optimus itself was developed, where the humanoid hardware was already defined.
Why a $100 Million Seed Round Stands Out
A seed round is typically a company's first major fundraising stage, often used to prove out an early product before a larger Series A. Seed rounds commonly range from a few hundred thousand dollars to several million; a figure near $100 million is far outside that norm, according to Cryptobriefing's coverage of the deal.
The scale of the round signals strong investor confidence in both Kumar's track record from Tesla and the broader bet that skilled-labor automation in manufacturing and aerospace is a large enough market to justify an outsized initial check. It also gives Intelligent Machines significant runway to hire engineers and, eventually, fund the hardware development its leaders have discussed.
What This Means for the Industrial Robotics Field
Intelligent Machines enters a market where most attention has gone to humanoid robots like Optimus, Figure's robots, and others aimed at general-purpose labor. By targeting specific skilled tasks in manufacturing and aerospace instead, the company is betting that narrower, task-built machines can reach commercial usefulness faster than general-purpose humanoids.
For readers tracking the robotics industry, the company is worth watching for two reasons: its founding team's direct experience building one of the most prominent humanoid robot programs, and its reinforcement-learning-first approach to teaching skilled physical tasks. Its hardware plans, and any named customers, have not yet been detailed publicly.
Frequently Asked Questions
Is Intelligent Machines building humanoid robots like Optimus?
No. It is building non-humanoid, task-specific robots aimed at skilled work in manufacturing and aerospace, a deliberate departure from the humanoid Optimus project Kumar led at Tesla.
Has Intelligent Machines released any hardware yet?
Not yet. The company is currently focused on software and reinforcement learning, though leaders have discussed designing their own hardware in the future.
