IRIS8 Talk to us

For the CIO

Who answers for the model in production?

IRIS8 does. The team that trains your model runs it in your environment, and you own it.

Your brief

What you answer for

  1. 01A model that works on the day it goes live and a year later
  2. 02A cost the board can plan
  3. 03A supplier you can leave without losing the work

Questions

Five questions CIOs ask us

The answers we give on the first call.

  1. 01

    Who is accountable when the model gets something wrong?

    IRIS8 is, for the model and its operation. The engineers who trained it run it, watch it and retrain it, so there is no handover between a builder and an operator. Your staff stay accountable for the business decision, because the model prepares work for a person to review.

  2. 02

    Do I need to hire an ML team?

    No. Your subject experts set the pass mark and review sample outputs. Your IT team grants access. IRIS8 does model selection, data preparation, training, benchmarking, deployment and daily operations. If you later want your own team to take part, they work in the same console we do.

  3. 03

    How long before something is in production, and what does it cost to find out?

    The pilot is fixed price: one use case, one production KPI, with a target of 8 weeks from kickoff to a model serving in production. The pilot fee is credited to year one. You pay for infrastructure directly, or it is passed through at cost.

  4. 04

    What do we own, and what happens if we part ways?

    You own the fine-tuned weights, the datasets, the benchmarks and every record. The model runs in your VPC or on-prem, so it stays where it is if we part ways. Nothing has to be exported or rebuilt for you to keep using it.

  5. 05

    We already pay for a frontier API. Does this replace it?

    Not all of it. Routine, high-volume work moves to your own model. Tasks that need a frontier model still go to one, through a gateway you control. There is no lock-in to one model vendor, and the calculator shows the cost on your own volume.

Who decides and who runs

You decideIRIS8 does
Which workflow goes firstScopes it and sizes the hardware
The pass markBuilds the benchmark and runs it
What data the model may seePrepares and masks that data
Whether a new version goes liveTrains, tests and stages the version

You keep: weights, datasets, benchmarks, records

Four decisions stay with you. The work is ours.

Next step

Bring one workflow to a 30-minute call