Home/Work
Selected work

Shipped to production, not slideware.

Engagements our principals scoped and shipped, anonymised and shown without client figures. The point is the pattern, and patterns travel well beyond where they started.

TypeProduction
SpanData · ML · BI
CloudsAWS · Azure
Pharma · Data
Data engineeringCloud

From scattered ERP data to one governed lake, with forecasting on top.

A manufacturer with source-of-truth data locked in its ERP, no central repository, and error-prone manual reporting. We built a direct ERP-to-cloud pipeline, a two-zone lakehouse with governance and lineage, and a demand-forecasting layer once the lake was live.

Outcome A working pipeline from source systems through to dashboards, stood up in weeks, with every layer in production and ready for AI on top.
Apparel · AI/ML
ForecastingReplenishment

Person-dependent ordering, replaced by forecasting wired into the system.

A complex SKU portfolio with heuristic allocation for new lines and manual daily replenishment. We built statistical demand sensing with custom regressors and closed the loop directly into the existing planning system.

Outcome Forecast accuracy lifted into the high-performing range, stockouts and excess inventory both down, manual ordering removed from the workflow.
Manufacturing · Shop floor
Operational intelligenceIoT

Real-time root-cause attribution across a large, multi-plant production network.

Static OEM dashboards, root causes for efficiency losses invisible at the system level, data scattered across sources. We built near-real-time line monitoring with custom machine states, root-cause attribution, and preventive-maintenance triggers.

Outcome Previously invisible root causes surfaced, machine downtime reduced, and reactive servicing replaced with statistical preventive triggers.
Manufacturing · Vision
Computer visionQuality

Defect classification at production speed, wrapping the cameras already on the line.

Existing inspection hardware caught defects at only the coarsest level; deeper classification was manual, leaky, and memory-constrained. We layered image models onto the current cameras via software, with no hardware to rip and replace.

Outcome Fully automated quality classification at line speed, manual inspection effort all but removed, and no capital spend on new equipment.
CPG · Multi-geography
ForecastingPlanning

Hierarchy-aware forecasting across regions and business lines.

Inconsistent, single-method forecasts that missed key demand drivers and lagged actual demand by a quarter. We formulated and tested demand-driver hypotheses with the business and built an ensemble engine that picks the best method per series.

Outcome Materially more accurate forecasts across geographies and lines, plus a reusable demand-driver library for future launches.
Logistics · BI
Data engineeringDecision boards

A sprawl of manual reports, collapsed into a handful of decision boards.

Daily dashboards taking hours to refresh, no central warehouse, and hundreds of overlapping manual reports with no single source of truth. We built a lakehouse, modernised ingestion, and replaced report sprawl with role-specific boards and exception alerts.

Outcome Time-to-truth cut from hours to minutes, manual reporting effort sharply reduced, and cloud spend optimised alongside it.

All engagements anonymised. Client names and figures intentionally omitted. Detailed, reference-backed write-ups available under NDA on request.

Why it travels

The pattern matters more than the industry.

Each engagement above started in one sector, but the shape of the problem repeats everywhere. A governed data layer, demand sensing, vision-on-existing-hardware, exception-based decision boards: the same patterns land in industries we'd never name in the same sentence.

A

Same data shapes

ERP, transactional, IoT, and document data look structurally alike across sectors, so ingestion, governance, and modelling patterns carry over.

B

Same decisions

Forecasting, root-cause, quality, and margin questions recur in different language. The maths and the architecture don't change with the logo.

C

Pattern recognition, fast

Because the same people who scoped these also build, they recognise the shape of a new problem fast, and skip relearning the basics on your time.

Where we fit

The honest version of who to hire.

Most firms pitch only their own strengths. Here's the comparison we'd write if we sat on your side of the table, including where someone else is the better call.

The alternative Where they shine Where we're the better call
Large system integratorsScale & outsourcing
Throughput at scale, multi-year managed services, full-stack IT outsourcing and vendor management.
When you want depth of thinking and direct accountability over headcount. Substance over slide volume.
Strategy consultanciesBoard-level framing
Transformation programmes, executive communication, and strategy for the boardroom.
When the answer has to be working software: code, pipelines, a live data layer, not a recommendation deck.
Large analytics / staffing firmsBig bench
Large dedicated teams running many parallel workstreams with mature global delivery.
When you want senior people personally accountable, with the engineering depth of those who scoped it also building it.
Freelancers & generalist agenciesFast & cheap
Low cost, quick starts, flexible hands for well-defined, lower-stakes tasks.
When the work is genuinely hard: production-grade, secure, and built to outlast the engagement.
OntegerBoutique studio
A small, senior team that consults and builds, end to end, from strategy through to production.
When senior people you can reach on a first-name basis, accountable for both the thinking and the build, is what the problem actually needs.

We're the right call when depth of thinking and direct accountability matter more than the size of the team or the logo on the contract. And we'll tell you plainly when someone else is the better fit.

Capability depth

One team, across the full stack of the work.

CloudsAWS · Azure · GCPCloud-native by default, aligned to whatever your stack already runs.
DataLakehouse & governanceSource-agnostic ingestion, lakehouse architecture, lineage, and warehousing.
AI / MLForecasting · vision · LLMsEnsemble forecasting, computer vision, NLP, and applied LLM and agentic systems.
EngineeringWeb · mobile · infraProducts, apps, APIs, CI/CD, and observable infrastructure built to scale.
Start a conversation

Let's build the thing you're stuck on.

Thirty minutes with a principal. No decks, no sales script, just a straight read on whether we are the right team to ship it.

Talk to our team Explore our services Or email hello@onteger.com