Services
One journey. The capabilities you need, where you need them.
Every engagement follows the same path, from the first question to AI running at scale. Join at the start, or at the stage you are already in.
Start, then all the way to scale.
- 01
Start Strategy and discovery
Where AI will actually move your numbers, what data you have, what it costs, and what to leave alone.
Outputs · Opportunity map, business case, roadmap - 02
Design Architecture and approach
Buy, prompt, fine-tune or build. Cloud, on-prem or on-device. Which agents, and which guardrails.
Outputs · Solution design, data plan, delivery plan - 03
Build Prototype to product
Data pipelines, models, agents and integrations into your tools, tested against real tasks.
Outputs · Working system, eval suite - 04
Deploy In the cloud and on the device
Ship to production wherever it needs to run, securely and within budget.
Outputs · Production release, runbook - 05
Operate Run it, measure it
Monitoring, evals, cost control and support, with our agents watching around the clock.
Outputs · Dashboards, monthly reviews - 06
Scale Grow what works
Roll out to more teams, add use cases, retrain and improve, and back to A.
Outputs · Scale plan, next roadmap
Plugged in where you need them.
Custom model training and fine-tuning
Your data, models you own, benchmarked before you commit.
On-device and edge AI
AI that runs in phones, cameras and machines, offline and private.
Agent systems
Agents that do real work inside your tools, with people in the loop.
Cloud and data platforms
The foundations everything else runs on.
Start
Strategy and discovery.
We start with the work, not a model. Which decision should get faster, cheaper, or newly possible? What data actually exists, and who is allowed to use it? What does a wrong answer cost?
We would rather say no. A checklist does not need a model. A pilot with nobody to own it will stall. Fine-tuning before a frontier API has been benchmarked is usually a waste. You leave with an opportunity map, a business case, and a roadmap you can start without us.
Deploy · in
On-device and edge AI.
In is on the device, at the edge, and on-prem. Out is the cloud. Deploy covers both, and some products need both: a small model in the camera, a larger one in the cloud when the network is there.
On-device work is a budget. Latency, power, memory, and privacy. We size the model to the device. You leave with a model that fits, a runtime, and a rollout plan.
Build
Custom model training and fine-tuning.
Renting a model forever is a fair default until it isn't. We benchmark the API first. If fine-tuning, or a smaller model you own, will win on cost, privacy, or quality, we train it and hand you the weights. If it won't, we say so before you pay for a dataset.
You leave with a dataset plan, the weights, and an eval report that says what improved and what did not.
Agents
Agent systems.
An agent is a worker with a narrow job, a set of tools, and a log. People set the bounds and accept the result. We don't ship an agent that can change production on its own.
The same pattern runs inside Atereta. Meet Scout, Forge, Relay and Ledger, and read the rules they work under.
Foundations
Cloud and data platforms.
Models and agents sit on something dull and necessary: data that is allowed to move, a place to run, and a bill someone can read. We put that in place as part of the journey, not as a separate programme of work.