All services
A spinning mill: rows of bobbins under a display showing the hall's temperature and humidity

Industrial IoT development

Plant and field data you can act on, because someone checked it first.

Industrial IoT fails quietly: sensors report, dashboards render, and nobody trusts the numbers enough to act on them.

Talk to our engineers →Follow one reading ↓

Follow one reading

Follow a reading from the device to a decision.

Field data arrives late, duplicated, gappy and occasionally lying, and analytics built on the assumption of clean input produce dashboards nobody trusts.

Simulated stream · …/env/temp

  1. 01 · Raw stream

    Field data arrives late, duplicated, gappy and occasionally lying.

    gap: no data, not zero99.90.099.9 rejected at the doorduplicate, kept oncelate, placed byits timestamp

    Hall temperature, average: 33.1 °C, what the dashboard shows

  2. 02 · Checked at the door

    Validated at ingestion, so data quality is caught at the topic level.

    gap: no data, not zero99.90.099.9 rejected at the doorduplicate, kept oncelate, placed byits timestamp

    Hall temperature, average: 31.7 °C, still pulled down by the zeros

  3. 03 · Gaps and duplicates named

    Gap and duplicate handling with explicit semantics.

    gap: no data, not zero99.90.099.9 rejected at the doorduplicate, kept oncelate, placed byits timestamp

    Hall temperature, average: 36.3 °C, computed from 41 readings

  4. 04 · Traceable

    Lineage, so a suspicious number can be traced to its source.

    gap: no data, not zero99.90.099.9 rejected at the doorduplicate, kept oncelate, placed byits timestamp

    Hall temperature, average: 36.3 °C, traced to its source

  5. 05 · Trusted

    Analytics built on data people have a reason to trust.

    gap: no data, not zero99.90.099.9 rejected at the doorduplicate, kept oncelate, placed byits timestamp

    Hall temperature, average: 36.3 °C, a number someone can act on

The namespace

The unified namespace, done properly

A unified namespace is the difference between a factory’s data being an asset and being an archaeology project: one coherent topic hierarchy, consistent payloads, every system publishing into a structure the next system can trust. The discipline is in enforcement, which is why we built and open-sourced a validator that checks topics and payloads against the declared structure, so data quality is caught at the topic level instead of discovered in the dashboard.

One topic hierarchy

enterprisesitearealinecellenv/tempenv/rhstate

Checked at the door

…/cell/env/temp

{ "value": 36.2, "unit": "°C", "ts": "…" }

accepted

…/cell/env/temp

{ "val": "36,2" }

rejected: "val" is not a declared field

…/cell/env/tmep

rejected: topic not in the namespace

Where the data comes from

Brownfield is the normal case.

Protocol bridging from legacy PLCs and serial equipment into a modern namespace is standard work, and our embedded practice handles the physical interfacing.

Protocols

MQTT and Sparkplug, OPC UA, OCPP, Modbus and the industrial serial family, plus the automotive CAN world through our gateway work. Protocol choice follows the equipment, not fashion.

From an assessment

What a plant’s data said when we first looked

For a discrete manufacturing plant we assessed six data files from its existing systems before designing anything: maintenance tickets, seam records, scheduling data, one welding station’s telemetry, the plan and the dispatch records.

01One welding station

A baseline from one station.

One welding station reported about every 31 seconds across 68 parameters. That gave the first per-seam arc-time baseline.

02The plan against dispatch

-60 days-40 days-20 daysplanned endabout 18 days67 days

The plan and the shop floor had drifted.

On the rows that matched, dispatch ran about 18 days ahead of the planned end, and one by 67 days. A buffer measure ran at 128 to 132 percent.

03Joining the plan to dispatch

The plan could not be joined to dispatch.

The two had no shared key. In one monthly plan only 7 of 182 rows matched a dispatch record, and the plan format had changed during the year.

04One month of maintenance

about 90 percent cleared in under four hours
about 81 percent electrical

Maintenance showed where the time goes.

One month of tickets held 359 entries. About 90 percent were cleared in under four hours and about 81 percent were electrical.

05 · What it changed

So the gate is calibrated to what the floor does.

Because the documented plan and the floor had drifted, the safety gate between agents and machines is calibrated against measured behavior before the design is frozen.

In the open

Published, so it can be checked.

EV charging is industrial IoT with a revenue meter attached, and OCPP is its lingua franca. We work across the protocol span: charge point management integration, telemetry capture, and the operational analytics above it.

Open source

A validator for topics and payloads

Open source
checks topics and payloads against the declared structure
22
passing tests in version 0.1
Live capture
verified against a real MQTT broker

Charging infrastructure

An open-source degradation model for EV chargers

OCPP
telemetry as the input
Open source
code and preprint public
Field validation
on an instrumented charger is under way

Tech4Biz works with Tier 1 enterprises. Client names stay private.

What it covers

From the topic hierarchy to the dashboard.

Edge-to-cloud platforms on open standards: OCPP for charging infrastructure, unified namespace architectures for manufacturing data, MQTT and industrial protocol integration, and the telemetry pipelines that make field data usable. Plus the analytics layer, built by people who have published degradation models against real telemetry.

How we engage

How the work is set up

Team
A dedicated team that steps into the chaos, takes ownership and ships, whether that’s one module or the whole product. Billed monthly, with a three-month minimum.
Starting
Within one to two weeks of agreeing what gets built, what gets left out, how payments work, and the roadmap after launch.
Working hours
7am to 8pm IST, with calls until 10pm IST. That covers the US Eastern morning.
Ownership
All code and IP are yours on payment.
Architecture only
We can design the architecture for a team you already have. Your team then owns delivery. We take responsibility for execution only when our own team builds it.

Questions

Before an engagement

Can you retrofit existing equipment?
Yes; brownfield is the normal case. Protocol bridging from legacy PLCs and serial equipment into a modern namespace is standard work, and our embedded practice handles the physical interfacing.
Which protocols do you support?
MQTT and Sparkplug, OPC UA, OCPP, Modbus and the industrial serial family, plus the automotive CAN world through our gateway work. Protocol choice follows the equipment, not fashion.
How is security handled?
Per-device identity, certificate-based provisioning, encrypted transport, and network segmentation designed with the same seriousness as the rest of our infrastructure practice.
Who owns the data and the platform?
You do. Open standards and open-source foundations are the anti-lock-in position, and the namespace design is documentation you keep.
Read the full explanation of the approach

Industrial IoT fails quietly: sensors report, dashboards render, and nobody trusts the numbers enough to act on them. The gap is engineering discipline between the edge and the decision.

What we do

Edge-to-cloud platforms on open standards: OCPP for charging infrastructure, unified namespace architectures for manufacturing data, MQTT and industrial protocol integration, and the telemetry pipelines that make field data usable. Plus the analytics layer, built by people who have published degradation models against real telemetry.

The unified namespace, done properly

A unified namespace is the difference between a factory’s data being an asset and being an archaeology project: one coherent topic hierarchy, consistent payloads, every system publishing into a structure the next system can trust. The discipline is in enforcement, which is why we built and open-sourced a validator that checks topics and payloads against the declared structure, so data quality is caught at the topic level instead of discovered in the dashboard.

Charging infrastructure and OCPP

EV charging is industrial IoT with a revenue meter attached, and OCPP is its lingua franca. We work across the protocol span: charge point management integration, telemetry capture, and the operational analytics above it. Our public work leads here: an open-source degradation model built on OCPP telemetry, with the methodology published and checkable and field validation under way.

Telemetry you can act on

Field data arrives late, duplicated, gappy and occasionally lying, and analytics built on the assumption of clean input produce dashboards nobody trusts. Our pipelines treat data quality as a first-class engineering problem: validation at ingestion, gap and duplicate handling with explicit semantics, lineage so a suspicious number can be traced to its source, and models validated against the ugly reality of production telemetry rather than a curated sample.

If nobody trusts your plant data enough to act on it, let’s talk.

Start the conversation →