BASED IN DELAWARE

A US company, built for the European manufacturing belt

Haltless, Inc. is registered in Delaware, United States. Delaware governing law and the Delaware courts apply to every customer contract, so you always know which courthouse will hear the case.

Customer telemetry stays in the United States. The full sub-processor list is set out in Annex III of our Data Processing Addendum.

The Story

Haltless was founded with a single premise: sensor data holds the secret to machine health, but traditional systems make that data too complex to decode in real time. We saw maintenance teams overwhelmed by false alarms and confusing dashboards, while critical failures still caught them by surprise.

We built Haltless to be different. A lightweight Python edge agent reads the PLCs you already have, and three deterministic anomaly detectors score every machine with a published formula instead of a black-box model. No data scientists, no six-month onboarding, no model drift to babysit. Deploy the agent on a Linux box near the machine and read the score on the page.

Haltless, Inc. is registered in Delaware, United States. The team writes Python, Go, and TypeScript, and ships from workstations that sit closer to a calibration bench than to a startup beanbag. Predictable, conservative, and uncomfortably honest about what the platform does and does not do.

WHO BUILDS IT

One founder

Attila Peter Szucs

Founder, CEO & CTO

The same person sets the product direction and owns the architecture underneath it. Ask a hard technical question during an evaluation and you get an answer from someone who can open the code, not a handoff to another department.

His background is electrical engineering, industrial automation, PLC programming and cyber security. Those four are the whole problem: read the controller correctly, do the arithmetic on top of it correctly, and put a box on an OT network without opening a hole in it.

Haltless is being built toward one commitment: no black box. You should be able to see which reading produced a score, which threshold moved it, and redo the arithmetic yourself when you don't believe it. A number you can't check is just a vendor asking you to take its word.

WHAT WE BELIEVE

Operating principles for industrial software

Four positions we are not willing to negotiate, because every one of them changes a decision the platform has to make.

Operators trust what they can reproduce

A black-box model that says "risk imminent" loses the room after the third false positive. We publish the formula. When a maintenance lead and an auditor look at the same score, they reach the same conclusion.

Plants have flaky networks

Half of competitor demos fall over the moment you yank the ethernet cable. We tested ours on real shop-floor networks with multi-day outages. The edge agent buffers in SQLite and keeps collecting.

The audit trail is the product

An HMAC-SHA256 hash chain, append-only database triggers, and seven years of retention. Compliance is not a feature we tack on later. It is the substrate every other feature stands on.

Per-machine pricing is honest pricing

You pay for what you monitor, not for seats or for cores we provisioned for ourselves. The number in the quote is the number on the invoice, with any applicable tax shown before you confirm rather than turning up afterwards.

HOW WE BUILD

Engineered for the realities of manufacturing

The four engineering decisions that shape every release. None of them were made to sound impressive on a slide.

Deterministic by default

Three classical anomaly detectors run in parallel: z-score against a rolling baseline, EWMA for gradual drift, and a rate-of-change check for velocity anomalies. No neural net, no training data, no model drift to babysit. The maths is on the page.

US-hosted by default

Production runs in the United States. Customer data residency is the start of the conversation, not an enterprise upsell. The sub-processor list is set out in Annex III of our Data Processing Addendum, with the actual processing region called out per provider.

Multi-tenant by Row-Level Security

PostgreSQL RLS pins every query to the current tenant. Not an ORM filter, not application middleware: enforcement happens at the database engine. A tenant administrator physically cannot reach another tenant's data.

One platform, not two stitched together

Alert, work order, parts, signoff, audit log: one tenant, one user table, one trail. The seam between telemetry and CMMS is the platform's strongest line, not the place where two SaaS bills meet through a Zapier flow.

Our Approach

Simplicity at the Edge

A small Linux agent sits next to your machines, normalises OPC UA, Modbus TCP, CSV, and JSON inputs, and forwards them to the Haltless cloud over TLS. A local SQLite buffer holds the data during WAN outages and flushes when connectivity returns, so nothing on the floor is lost.

Actionable Intelligence

An alert is only useful if it tells you what to do. Every Haltless health score traces back to the specific event and the specific signal that moved, with a formula your auditors can read.

HONESTY GUARDRAILS

What Haltless is not

Industrial software sells better when it claims to do everything. We have found the opposite to be true with the buyers we want. Here is what we are explicitly not building.

We are not a SCADA or DCS

Haltless complements your supervisory control system. We do not write to PLCs, we do not close control loops, and we do not coordinate emergency-stop arrangements. If your buyer is asking about real-time control, the answer is your existing SCADA, not us.

We are not a real-time control system

Alerts may be delayed, missed, or wrong. The Services are not a safety-instrumented system within the meaning of IEC 61511, and they are not certified to any functional-safety standard. Customers must not rely solely on Haltless for safety-critical control decisions.

We are not a black-box AI model

There is no neural network in the prediction service. The health score is a deterministic formula with a published deduction table. If your team cannot reproduce a score on paper, the platform has lost its purpose.

We are not a sensor reseller

If you do not have OPC UA, Modbus TCP, or a CSV / JSON drop in place, we will say so on the call and point you at the right hardware vendor. We do not bundle proprietary sensors you do not need so we can show a hardware line on your quote.

Join the Shift

We are empowering the next generation of smart manufacturing. Stop reacting to failures and start predicting them.

Request a quote

COMMON QUESTIONS

Questions about the company

Who is behind Haltless?

One founder. Attila Peter Szucs is Founder, CEO and CTO, with a background in electrical engineering, industrial automation, PLC programming and cyber security. Haltless, Inc. was founded in 2026 and is registered in Delaware, United States.

How big is the company?

Small enough that the person setting product direction also owns the architecture. That is stated plainly rather than hidden behind a team page, because it changes what an evaluation looks like: a hard technical question gets answered rather than routed.

Is Haltless funded, and does it have customers?

Haltless is pre-launch with general availability scheduled for Q1 2027, so there are no customers yet and no case studies to cite. Any vendor at this stage offering references is worth a second look.

Why not use machine learning?

Because a maintenance score you cannot reproduce is a score you cannot defend. Three deterministic detectors give the same output for the same inputs every time, the formula is published, and every score is recomputable on paper by your own engineer or an auditor.