Liquid Electronics — AI Services & Hardware Engineering

Custom AI systems
Built for your business
Technology you own

We build AI systems for small businesses—self-improving systems you own and control. We’ll teach you how to maintain and expand your own systems, or you can choose an ongoing support plan.

Be independent

Software or hardware, custom tailored for your business. Own the technology you use. Change it whenever you want.

Be free

Portable by design. Take your context with you. Switch hardware or cloud platforms whenever you want.

Choose your AI

Model-agnostic by design. Switch AI providers at any time, or run open-weight models on your own hardware.

Control your data

Keep sensitive data where you want it. You decide what AI can see and what stays off-limits.

Forward Deployed Engineering

Our approach puts an engineer alongside your business—from understanding your goals through implementing polished systems. We lead the engineering, you direct the results. We’ll teach you how to maintain and expand your own systems, or you can choose an ongoing support plan.

Understand your business.

We’ll start with a short consultation. Tell us about your business, ask questions, and explore where AI could make a useful difference. Together, we’ll see whether we’re a good fit.

If we decide to work together, we take a closer look at your goals, existing tools, and day-to-day work. You bring the business knowledge and real examples; we identify where connecting your tools or building something new could help.

Agree on the results.

Together, we choose a useful first workflow and define what success looks like: time saved, fewer errors, or another result that matters to you. Before building, we agree on what we’ll deliver, what it will cost, and how we’ll test it.

We also discuss what you want to learn and how involved you’d like to be. Together, we decide who will look after the system and what support you need.

Build around your work.

We connect your tools and information, build the missing pieces, and put AI to work where it helps. We develop the system in working versions, so you can try it with familiar tasks and give feedback as it takes shape. That feedback helps us refine how it fits into your day-to-day work.

We build in your access rules and approval steps. You can see what the system does. Decisions that need your judgment stay with you.

Test it. Put it to work.

We test the system with real examples from your business, including missing information, unusual situations, and cases that need your attention. We check that it meets the requirements we agreed on in step 2.

Before launch, we show you how to recognize a problem and get help. You’ll have clear instructions for carrying on with essential tasks if the system goes down.

Own it. Understand it. Keep improving it.

Learning is part of delivery. We walk through the system with you and make sure you understand it. We also make sure you have the resources needed so that you or any software engineer of your choosing can improve or expand on it, knowing that you can always return to a previous version if something breaks.

You receive the source code, settings, and a practical guide to running your system. The guide explains which services your system uses, such as AI providers or other software subscriptions. It lists what each costs, and where your billing and account settings exist. Third-party software, models, and services keep their own licensing and terms.

Ongoing support from Liquid is always an option. We can train your team, answer questions as they come up, and help you plan your next improvements. Whether you want to add a new feature, connect a new tool, or build another workflow, we’re here to help as your needs change.

Use cases

Retail

Turn hours of work into a few text messages

Once a week, a salon owner checks inventory, writes down what’s running low, compares supplier prices, and places orders. But a shampoo that usually sells slowly unexpectedly sells out at the beginning of the week. The shelf stays empty until she notices during her next inventory check, places an order, and the shipment arrives. A client who would have bought it leaves without it, directly affecting her bottom line.

We build a system that uses the cameras she already has to monitor her inventory throughout the day. It tracks outstanding orders and orders replenishments before stock runs out, comparing approved suppliers by total cost, including shipping, and delivery time.

Because she owns the system, she can have it extended with new connections and rules as her needs change. For example, connecting her calendar lets the system automatically plan around her vacations. Two weeks before she leaves, it reviews past sales for the same time of year and recent demand to forecast how much of each product her team will need while she’s away. It accounts for stock on hand and incoming deliveries, then orders the additional quantities within her spending limits.

The system compares sales, profit margins, and unsold stock to recommend products to stock more of, replace, or try. She chooses the changes. It measures the results and uses them to refine its next recommendation.

She sets the priorities and spending limits. The system handles the routine work and brings decisions outside those limits back to her.

She’s on her feet with clients all day, so a computer interface isn’t ideal for her. Instead, the system texts her from its own dedicated phone number with specific questions or decisions that need her attention. She replies directly, and the system carries out the approved action.

For her, hours of inventory and ordering work each week collapse into a few targeted questions sent by text to her phone, freeing her time and attention for clients, growing her business, and pursuing new opportunities.

Illustrative example of a system we could build around your business.

Restaurant group

Scalable solutions built into every location

Consider a group of twenty restaurants. At the published prices used here, subscriptions for inventory management, seven kitchen displays, and three digital menu boards at each location total $111,600 a year.123

In this scenario, Liquid starts with a restaurant platform that has already been developed. We adapt it to the group’s needs, reusing compatible equipment. The group pays for that development work once, owns the resulting system, and shares the benefits across all twenty restaurants.

Customization and central computing cost $120,700 before tax. With eight hours of IT support a week, electricity, equipment replacement funding, and an illustrative $200 a month for cloud AI, modeled ongoing costs total about $61,500 a year. Compared with the selected subscriptions, that leaves about $50,100 a year, with simple payback of about 2.4 years.

The shared system can evolve with the business, building on what it learns, improving over time. Each improvement can benefit every location, making the system more useful across the whole group.

Costs

Current kitchen display subscriptions$33,600/yr
Current inventory subscriptions$71,520/yr
Current digital signage subscriptions$6,480/yr
Current subscription costs$111,600/yr
Customization and implementation$105,500
Two AI computers — ASUS GX10 · Hardware sizing $11,998
Four 12 TB backup drives$2,796
Two external backup-drive enclosures$400
Initial costs$120,694
IT support budget — 8 hours/week$55,009/yr
Electricity — modeled load$1,006/yr
Central-equipment replacement fund — five-year funding$3,039/yr
Cloud AI for software improvements — illustrative $200/month$2,400/yr
Modeled annual costs$61,453/yr
Modeled annual savings$50,147/yr
Simple payback2.4 years

The cloud allowance covers all twenty locations and is an assumption, not measured usage. Actual charges depend on the models and workload; any additional paid tools or services would increase costs. These are illustrative costs and savings, not a customer quote.

More than subscription savings

A kiosk fails during dinner. The system detects the failure, reserves a compatible spare at another restaurant, and adds it to the loading list for the next scheduled transfer. The spare travels with the regular supplies on an existing route. A small status ribbon on the working kiosk shows the replacement’s progress: reserved, loaded, en route, and received. The team stays informed while focusing on the dinner rush. The manager doesn’t have to coordinate the replacement.

A regular catering customer increases today’s order from sixty people to a hundred. Normally, the restaurant could handle it. But the system calculates capacity from current staffing and kitchen workload. It knows one of the prep cooks hasn’t clocked in, sees what orders are already queued, and forecasts incoming orders using historical data. Comparing that workload with the available staff-hours and kitchen resources, it determines that the restaurant is already short of the capacity needed to keep orders on time. Adding the extra portions would increase customer wait times.

The closest restaurant is also at capacity, so the system finds a third that can prepare the extra forty portions. That covers the increase, but leaves the first kitchen overloaded. Knowing preparation hasn’t started, the system checks whether the third restaurant can take all hundred and finish early enough to allow for the longer drive.

The system transfers the order between restaurants and adjusts its fire time to meet the original delivery time. The customer gets their order on time, while the reduced workload helps the short-handed team bring ticket times back on target. Each kitchen sees the updated order and timing in its queue. The managers don’t have to coordinate the transfer.

In the group’s own software, a common substitution takes several screens and sends the kitchen an unclear ticket. A staff member records a short video, explains the frustration, and sends it through the staff app. Similar reports from other locations reveal a recurring problem.

The system groups the feedback, assesses its impact, and turns it into a proposed improvement. Development agents use the original videos and messages to guide how they build and test the proposed changes. Staff and managers don’t need to write code or manage development; they review the proposed solutions and accept or reject them.

Employees’ firsthand experience becomes a valuable asset. Their suggestions improve the business while removing frustrations from their own day. They gain a real say in the tools and processes they use, and a reason to keep contributing: their ideas can make work better for themselves and their colleagues.

The system measures whether each change helps. For the substitution fix, it compares completion times and ticket errors before and after the change. Those results guide whether to keep, adjust, or reverse the change. The system remembers what works and uses that knowledge when developing future improvements—turning employee feedback into an automated cycle of better tools and smoother operations.

Over time, that knowledge becomes a business asset the group can keep building on, with potential to train or fine-tune its own specialized models or license the data for AI training.

Connecting your own intelligence to your own data allows you to spot patterns you would never have seen before and forecast more accurately, with less effort than ever before.4

AI on your own hardware

Tell us about your business and what you’d like to improve.