Skip to content
Part of Dynamis Group
Client Portal

Feasibility studies, dataset creation, scoped proofs of concept.

Two-to-four-week engagements that establish whether a perception, classification, segmentation or re-identification task is tractable on your data — with a written go / no-go either way.

Dynamis Labs — Prototyping is the early-stage engagement pillar at Dynamis Group: scoped two-to-four-week feasibility studies, dataset creation and validation (Label Studio, CVAT, Cohen's kappa for inter-annotator agreement), and early-stage system design. Output: data audit, baseline benchmark on a task-appropriate model architecture, reproducible notebook, and a written go / no-go memo. Pricing is fixed-fee on defined scope.

  1. Research
  2. Prototyping
  3. Production
  4. Licensing

Workstreams

What gets built before anything ships.

Three shapes of prototyping work, each scoped to a written outcome. None of them require a commitment to the production engagement that may follow — the answer can always be "don’t".

Feasibility studies

A two-to-four-week scoped study to establish whether a perception, classification, segmentation or re-identification task is tractable on your data, your operating constraints and your budget. The output is a written go / no-go either way — with the baseline benchmark, the data audit, and the reproducible notebook attached.

Dataset creation & validation

Where the off-the-shelf dataset doesn’t cover the problem, we build one. Sampling strategy, annotation schema, inter-annotator agreement, lineage and consent documentation. Validated to the standard a downstream evaluation can stand on.

Early-stage system design

The architecture sketch before anyone trains a model: data ingress, augmentation strategy, evaluation methodology, deployment topology, and the failure-mode register. Written as a working document the production team will inherit.

Where a prototype clears its go / no-go and moves to a six-to-sixteen-week build, the work continues at Dynamis Labs — Production.

Common questions

FAQs

Here are some of our most frequently asked questions. Can't find what you're looking for? Reach out to our support team.

What is a feasibility study?
A scoped two-to-four-week engagement that establishes whether a perception, classification, segmentation or re-identification task is tractable on the client's data, operating constraints and budget. Output: data audit, a baseline benchmark using a model architecture chosen to fit the task, a reproducible Jupyter notebook, and a written go / no-go memo. The answer can — and sometimes should — be "no".
What does early-stage system design produce?
A working architecture document for the proposed CV system: data ingress, augmentation strategy, model selection rationale, evaluation methodology, deployment topology (edge / cloud / hybrid), failure-mode register, and a runbook outline. Written so the production team that inherits it can ship from the document — not a sketch, not a slide deck.
What does a feasibility study cost?
Quoted per engagement based on data complexity, annotation requirement and team scope. Two-to-four-week feasibility studies typically run between AUD $20k and $80k inclusive of the written go / no-go memo and reproducible notebook. Larger dataset-creation engagements price separately. Pricing is fixed-fee on a defined scope, not time-and-materials open-ended.
Do you create datasets?
Yes, where off-the-shelf datasets do not cover the problem. Sampling strategy, annotation schema (often via Label Studio or CVAT), inter-annotator agreement metrics (Cohen's kappa, Fleiss' kappa), lineage and consent documentation. Validated to a standard a downstream evaluation can defend — including audit-grade if the deployment will be regulated.
Is the prototype reusable in production?
Sometimes — the baseline notebook and the dataset transfer cleanly. The prototype model is rarely production-ready: production engineering (evaluation harness, inference service, deployment artefacts, drift monitoring) is what the Production engagement adds. Prototyping proves the question is worth asking; Production answers it at scale.

Start a conversation

One architect, one inbox.

Bring us the situation. We’ll pair you with a solution architect and write back — no hand-offs across divisions, no sales cadence.

Get in touch