An example workflow
A test prep app
Doubt requests spike in the weeks before exams. An evaluation could run a sample of past doubts with approved answers through right-sized models and compare accuracy, latency and total cost.
Your learning features, tested on your own evals.
Doubt solving, practice questions and test prep, tested on right-sized models against your evals and your teachers' review.
Prefer email? hello@azitalabs.com, subject “Workload evaluation: Edtech & coaching”
Exam season arrives on schedule. The answers have to be right.
Every student question calls a model, and a simple doubt that a small model answers correctly still runs on a frontier one. Usage peaks in exam season, answers have to meet your teachers' standard, and students feel every delay. You cannot tune what you rent, and rate limits arrive on someone else's schedule.
- What you bring
- A set of real student questions with teacher-approved answers, and a month of usage.
- What comes back
- Answers, explanations and practice questions scored on your evals, for your teachers to review.
What an evaluation could look like.
An example workflow
A coaching institute
Teachers write practice questions for every chapter. An evaluation could take one chapter's approved questions, generate new ones, and let the faculty compare them and the total cost.
An example workflow
A language learning platform
Speaking and writing feedback must be accurate and encouraging. An evaluation could run a sample of learner responses through the team's eval set and compare feedback quality and total cost with the current provider.
Your evals decide.
Your evals are the bar
Every step is judged against your teachers' evals. A step that does not pass stays where it is.
Your teachers review the answers
Answers and explanations are scored on your evals and reviewed by your teachers before any student sees them.
Quality and total cost, compared
Each learning feature's total cost compared with what you pay today, before exam season arrives.
Where it runs, and what is agreed.
Capacity from the Azita fleet
Run your workload on capacity from Azita’s installations. Scope the workload and data requirements.
Discuss your workload on the fleetAn installation in your building
Scope the room and workload together. Azita designs, installs and commissions the deployment.
Discuss an installation in your buildingWhere the work runs, which models are used, who has access and how long anything is kept are agreed in your proposal before any work starts. The Trust page sets out custody and residency by side.
Compare quality and total cost on your own workload.
Agree a representative job and the acceptance criteria. Compare the output, turnaround, data requirements and total cost before deciding.
- Share one workflowOne learning feature in production, with its eval set and last month's usage.
- Agree the pass criteriaAnswer accuracy your teachers accept, the response time students feel and your content standards.
- Compare on both stacksPass rates, latency, data requirements and total cost, side by side. Cut over step by step, or not at all.
Available to evaluate on your own workload.
On the Azita fleet, or on an installation in your building.
Arrange a workload evaluation
We'll review what you need to run and contact you to discuss the right setup.
Proposal based on your workload and deployment.
There is no price list. Your proposal states what is being quoted, the commitment and what is included. Human review, software licences and integration are listed wherever they apply.
Add what you pay for today. It travels with your enquiry as a note, so the comparison starts from the same work.
Prefer to read first? How pricing works
Straight answers.
Will our evals still pass?
Your eval suite decides. We map each step to the smallest model that meets your pass criteria and run the suite on both stacks side by side. A step that does not pass stays where it is.
How does an evaluation work?
Send one learning feature with its eval set and last month's usage. We agree the pass criteria, run the suite on both stacks, and set out pass rates, latency and total cost for each step.
What changes in our code?
Your evals run in CI, and you cut over step by step behind your own flags. Any step that does not pass on the new stack stays with your current provider.
Where does student data go?
On the Azita fleet, or on an installation in your own building. Where your student data is processed, who has access and how long anything is kept are agreed in your proposal before any work starts.
How is it priced?
Proposal based on your workload and deployment. It states what is being quoted, the commitment and what is included, and lists human review, software licences and integration wherever they apply. There is no price list.
What happens if we leave?
Term, exit and what happens to your student data at the end are set out in your agreement before you sign.
Related
Engineering and product
Agents, support, extraction and product features, tested on your own evals.
The page for every kind of work in this group, and how an evaluation works.
Read
Also for
Course creator
Subtitles, dubbed lessons, quizzes and notes, evaluated on a lesson you already published. Compare the output and total cost before you switch.
Read
Also for
Chatbots & support AI
Intent, retrieval and response steps on right-sized models, tested on your own support evals. Compare resolution quality, latency and total cost first.
Read
Start with your own workload.
Share one pipeline
A production workflow and its eval set. We agree the pass criteria, then compare output, latency and total cost on the same input.
Arrange a workload evaluation