ItInsightsWorld Startups Brief: CopperPilot's $390,000 Pre-seed Bet on an AI PCB Design Agent

If you are an early-stage founder weighing a move into hardware-tooling AI, the useful part of this announcement is not the size of the round but the three assumptions hiding underneath it: engineers will not migrate off their layout tools, an agent earns trust through verification rather than generation, and a usage curve is not yet a revenue curve. The hardware-design startup CopperPilot has raised $390,000 at pre-seed, with Neon Fund leading and U.S.-based angel investors joining, and the timing is deliberate — version 1 of its AI design agent, codenamed Epsilon, is opening to hardware engineers. As evidence of pull, the company points to 25-fold growth in weekly active users since August 2026 and more than 6,000 design tasks, logged internally as artifacts, run on its platform.

Deal snapshot: what was funded, and what was left out

Item	What the company disclosed
Round	$390,000 pre-seed; Neon Fund led, U.S.-based angel investors participated
Milestone	Opening of version 1 of Epsilon to hardware engineers
Traction	25-fold growth in weekly active users since August 2026; more than 6,000 design tasks run to date
Founded	October 2025, by Aditya NG, Mohit Bagur and Anish Harsha
Footprint	Dual headquarters in Bengaluru, India, and Houston, Texas
Use of proceeds	Product research and engineering, agent speed and reliability, engineering-team expansion
Absent	Revenue figure; user-to-paid-customer conversion rate

CopperPilot released neither a revenue figure nor a user-to-paid-customer conversion rate, so the traction numbers should be read as adoption evidence rather than commercial proof. Spending priorities, per the company, are product research and engineering, a faster and more dependable agent, and a bigger engineering team.

Origin: a team that kept hitting the same wall

From left: Mohit Bagur, Anish Harsha, and Aditya NG, the company's co-founders.

Aditya NG, Mohit Bagur and Anish Harsha founded CopperPilot in October 2025 around a blunt stated goal — “no hardware team is ever blocked because a board is too hard to design.” What they are building is an agent that absorbs part of the PCB design process: interpreting requirements, selecting components, drawing schematics, routing boards, and checking designs before manufacturing.

In an exclusive interview with EE Times, co-founder and CEO Aditya NG traced the idea to his own hardware work, autonomous vehicles included: “On almost every autonomous vehicle team I worked with, someone would eventually say it: A compute unit for six cameras is too hard to design, or a drive-by-wire unit is too much for us. Many teams stopped right there. The ones that kept going spent months staring at datasheets and paying for mistakes on board re-spins,” he said. “I built CopperPilot so that no hardware team is ever blocked because a board is too hard to design,” he added.

Adoption wedge: work inside the toolchain, do not replace it

Rather than asking teams to move designs into a new environment, the startup automates workflows inside the EDA tools engineers already run. KiCad, Altium and LTspice are supported, with the deepest integration centred on PCB design and firmware flows. Epsilon is also reachable through AI coding agents — Claude Code, Cursor and Codex — so a developer can hand board-design tasks over without leaving that editor.

The task chain the agent covers

1. Parses plain-English design requirements.
2. Reads component datasheets and manufacturer specifications.
3. Identifies suitable components and presents alternatives for the engineer to choose.
4. Creates schematics plus custom symbols or footprints.
5. Proceeds to PCB placement and routing.
6. Checks the design for electrical rule check (ERC) and design rule check (DRC) violations and manufacturing constraints.
7. Hands back a review report that ranks what to fix first.

Where requirements are incomplete or a component cannot meet a design constraint, Aditya said the agent asks for clarification instead of guessing, which keeps architectural judgement with the engineer. Where requirements are clear enough, it can run a sequence of design tasks autonomously, while engineers retain the option to review or intervene at individual stages. Interfacing with EDA tools mixes official application programming interfaces (APIs), direct file editing and other techniques. “The method varies according to what works for a particular tool or task,” Aditya said.

Cross-tool workflows are the next layer. In ordinary practice an engineer inspects firmware code, PCB design and communication protocols separately; by connecting those pieces, the company wants to surface interactions and help engineers reason through problems that span hardware and software. The added value, according to Aditya, is not only analysing a circuit board or writing firmware, but considering how different parts of an electronic system affect one another.

Verification, not generation, is the hard part

A schematic or layout is only half a design: it must also satisfy electrical requirements and the constraints of the manufacturer that will build it. ERC and DRC set a baseline, Aditya argued, but they cannot catch every problem, and his simplest example makes the gap obvious.

“While connecting two computers or two microcontrollers over a serial connection, you are supposed to connect the transmitter of one device to the receiver of the other, and the receiver of the first device to the transmitter of the other,” he explained. “It is very easy to reverse these connections, and it happens to many engineers. This is where CopperPilot really shines. It can catch mistakes that ERC and DRC checks cannot detect.”

Flagging is not the same as prioritising. “For engineers preparing a design for manufacturing, the agent can produce a review report that separates blockers from non-blocking issues, prioritizes fixes, and explains the remaining problems,” Aditya said. “It may also identify design changes that could reduce cost.”

Because hardware design has no single equivalent of a software compiler or linter that can confirm an entire design is correct, the startup built its own testing infrastructure. “With CopperPilot, we have internally built out our own harness, benchmarks, and tooling to validate the design,” he said. “The agent can read the datasheet, manufacturer specifications, and design tools, with the startup testing how well this process works through its harness and multi-agent framework.” Manufacturer-specific test suites, maintained component information and catalogued examples of design failures feed the same loop, so the team's understanding of how generated designs can fail grows as the agent is used.

Figures shared by Aditya: the system was evaluated against approximately 49 test cases based on human-designed circuits, with an 18% task completion rate reported for a baseline system and 78% for CopperPilot. Read that the way you would read any early hardware-AI benchmark — the baseline system was not specified, and the result comes from a small, company-run benchmark that has not been independently verified. The startup also publishes a new open-source board design each day, around 30 designs so far, inspectable in KiCad, which is the more auditable artifact of the two.

Who pays, and where those buyers are

As a B2B software-as-a-service (SaaS) company, CopperPilot sells across individual, team and enterprise tiers. Individual users get a free Starter plan plus paid Pro, Pro Plus and Ultra plans; paid team plans exist, and enterprise customers talk to the company about specific requirements and internal deployments. For those enterprise buyers the pitch is context: integration with existing tools and workflows, connections to GitHub and ticketing systems to draw on organisational knowledge, and a reusable skill an engineer can author to check a design against company manufacturing standards and share with colleagues — accumulated engineering experience folded into the agent's workflow.

On data, the company says that under its standard individual and paid plans, design data and chat history are stored locally on the user's computer, while information is sent to servers and a large language model (LLM) for processing. “CopperPilot has a zero data retention policy after an interaction,” Aditya said at the time of the interview. “We also offer on-premises hosting as an option, although it has not yet been deployed for a customer.”

Served audiences and target markets include individual engineers, startups, design service providers and larger businesses in IoT, health and medical technology, battery technology, electric vehicles and other automotive applications, with consumer electronics, renewables, energy storage and robotics named as further fits. Customers sit in the U.S., U.K. and Europe, users are also present in India and other countries, and the company maintains dual headquarters in Bengaluru, India, and Houston, Texas.

Common misconceptions this round invites

• Generation speed equals design quality. The company's own framing treats ERC and DRC as a baseline, with the serial-connection reversal as proof that the valuable step is catching what those checks miss.
• Usage growth equals revenue. No revenue figure or user-to-paid-customer conversion rate was disclosed alongside the raise.
• Local files mean nothing leaves the machine. Design data and chat history stay on the user's computer, but information is sent to servers and an LLM, and on-premises hosting has not yet been deployed for a customer.
• A completion-rate number speaks for itself. The 18% versus 78% comparison rests on approximately 49 test cases, a self-built harness and an unspecified baseline.
• An agent must replace the toolchain. CopperPilot's route is automation inside KiCad, Altium and LTspice, plus hand-offs from Claude Code, Cursor and Codex.

Action recommendations for teams reading this space

1. Founders building near EDA tools: treat integration as the product, not as a connector bolted on afterwards. Map which official APIs exist for each target tool, where direct file editing is the only viable path, and how you will switch method per tool and task.
2. Founders selling to hardware teams: design the blocker-versus-non-blocking review report before polishing generation quality, because that report is what a manufacturing hand-off actually consumes.
3. Investors sizing the round: the disclosed gap is commercial rather than technical, so push for revenue and paid-conversion data sitting behind the 25-fold weekly-active-user claim.
4. Enterprise evaluators: make the first diligence question about what leaves the local machine during processing, and demand a production reference for any on-premises promise.
5. Practising engineers: run one small board through the free Starter tier and compare its review report against your own ERC and DRC output; the daily open-source boards, around 30 designs published so far, let you audit builds without a sales call.

FAQ

What was raised, and by whom?

$390,000 at pre-seed, led by Neon Fund with U.S.-based angel investors joining, announced as version 1 of Epsilon opens to hardware engineers.

What traction did the company publish?

25-fold growth in weekly active users since August 2026, and more than 6,000 design tasks run on the platform. No revenue figure or user-to-paid-customer conversion rate was disclosed.

How capable is the agent, on the company's own numbers?

78% task completion against an 18% baseline, over approximately 49 test cases based on human-designed circuits — small, company-run and not independently verified. “Today, CopperPilot designs an AI camera board in about 30 minutes,” Aditya said.

Which tools and workflows does it touch?

KiCad, Altium and LTspice inside the existing EDA flow; Claude Code, Cursor and Codex as entry points; GitHub and ticketing systems for enterprise context.

Who is it built for?

Individual engineers, startups, design service providers and larger businesses working on IoT, health and medical technology, battery technology, electric vehicles and other automotive applications — with customers in the U.S., U.K. and Europe, users in India and elsewhere, and a company operating from dual headquarters in Bengaluru and Houston, Texas.

What is the next hurdle?

Making the agent reliable enough for engineers to use day to day, while adoption expands among individuals, startups and larger design organisations, with a shorter path from idea to manufacturable electronic product as the longer-term goal.

For a Startups reader, the read-through is narrow but instructive: the money is small, the wedge is a specific pain Aditya NG watched stop teams on autonomous-vehicle programs, and the durable claim is verification rather than speed. The signals worth tracking are whether weekly active users convert into paying teams, whether the internal harness ever faces an outside benchmark, and whether the on-premises option gets a first deployment. Aditya's own read on the endpoint is unguarded: “The racecars of the future will be designed with CopperPilot.”