Business

Seimeido (Sakai, Osaka) is a sole proprietorship that takes on chemical bath management for plating and cleaning lines, measurement and logging automation for experiments and inspections, and the design and prototyping of small purpose-built machines, for engineers in factories and research labs. We combine hands-on chemistry with machine learning, image processing, simulation, statistics, and data science. We build our own instruments and file our own patent applications. This page covers the work we take on, our inventions and patent applications, and the themes in progress, in that order.

Work we take on

Chemical Analysis & Research

On plating and cleaning lines, we start from deciding what to measure, by which method, and how the test conditions are set. Whether a value is measurement error or a change in the process cannot be told until the points accumulate. Control limits need about twenty analyses taken under the same conditions before they can be trusted.

Analytical values, replenishment history, test conditions, and appearance photographs tend to sit in separate places, so we bring them into one ledger. The accumulated values go through statistical processing and machine learning, and come back as a report saying whether to replenish, re-measure, or stop. Together with automated titration, we are working toward monitoring the bath in real time.

Experimental & Inspection Automation

We automate measurement and logging in lab experiments and on factory inspection lines. Sensors, fluid control, and instrument integration capture the data, and we design the work procedure into one continuous flow. Instrument integration on the factory side overlaps with the field called FA (factory automation); making equipment and its environment measurable overlaps with the field called physical AI.

Bath management on a plating line is still usually done by titrating by hand. Automate the titration and the number of measurements can rise without the labour rising with it. Statistical processing needs dozens of tests run under the same conditions. That count is hard to reach by hand, and the conditions drift along the way. When something goes wrong with a bath, statistical processing of the accumulated values also supports working out what changed.

Equipment built to be read off a dial by a person emits nothing outward. We add sensors or a link so it speaks in a form a machine can read. Where it is broken, we repair it, as long as it is small electronics. One data logger has gone back into service that way. Where a small dedicated machine is needed, we take it from circuit board design and firmware through to a prototype.

In development: a portable fully automatic titrator meant to work in developing countries.

Simulation & Visualization

We provide simulation and visualization of physical phenomena, mechanisms, control, and fluids in a form usable for design study and comparison. Where the number of conditions to sweep is large, machine learning narrows the search range and the results are assessed statistically.

Heat conduction and cooling, and the conductive and thermal paths that particles form (percolation), are within what we simulate. We take on consulting, and the examination of patent claims, on that basis. The results are checked against published values, and the performance figures of a material itself are settled against measurement. The in silico work used to design chemical formulations is described under themes in progress, below.

Proof of Concept

Before a full rollout, we take on the stage that establishes how much the data already in hand can tell you. What is visible, what is not, and why a call cannot be made are separated first, and that decides the conditions for going further.

On a wet process watched by camera, for example, we take the video and images already recorded together with the process operation log. What you want to see — foam, scum, deposits, stalled flow — and what the imaging itself contributes — reflections, steam, droplets, a dirty lens — are measured as separate indicators. Because the imaging disturbance carries a number of its own, a stretch that did not show up comes back as "cannot be judged" and stays distinguishable afterwards.

What you get back is an observability assessment, candidate features, the conditions that are hard to judge, the data still worth capturing, and the design of the next stage. Training the production model and building its accuracy, controlling real equipment, and estimating physical properties such as viscosity are the next stage's work, after observability has been established. This separation goes faster when it is read by someone who has run the process itself. What the screen is showing, in terms of the state of the process, is decided together with the person in charge.

Development Work & Overseas Coordination

We take on small purpose-built machines such as bag deaeration sealers as one job: mechanical design, circuit board design, firmware, parts sourcing, prototyping, and condition logging. Technical prototypes and AI / Web / IoT implementation come under the same heading. With overseas manufacturers in China and Southeast Asia, we help with specifications, documents, inquiries, and checks before introduction.

See our equipment manufacturing

Inventions and patent applications

We write and file our own patent applications, from the specification to the filing procedure. As of September 2026 we have filed four applications, in the fields of sintered-resin manufacturing processes, laboratory apparatus, optical components, and sensors. Titles and contents stay undisclosed until the statutory publication date.

Themes in progress

An angle sensor for agricultural concentrating solar towers, in progress as a development collaboration with NT Giken Industries. Business alliance announcement
We are working on water treatment for developing countries. The proof-of-concept design began as wastewater monitoring: how to tell foam, the early signs of clogging and a drop in flow apart from one another, generalised from there to wet processes at large. This is a return to where it started.
Lab automation, including a portable fully automatic titrator for use in developing countries, in development.
Simulation for design. As a tool for designing chemical formulations such as plating brighteners, we handle molecules at the electrode surface in silico, combined with machine-learning condition search.

Contact

From early-stage ideas to implementation planning, feel free to reach out.

Get in touch