Model training and fine-tuning
Tell us about the model, dataset and memory needs.
COMPUTE, WITH A HUMAN IN THE LOOP
On-demand computing, from GPU jobs to virtual workstations.
Describe the workload. We pick the right machine and reply with a setup and a quote.
Request computeNo commitment. Nothing charged by this site.
GPU jobs / CPU compute / Remote workstations
GPU sizer
Enter the model size and precision. We estimate the memory with two published rules of thumb and suggest a GPU memory class, then you can drop the result into your compute brief.
01 / HOW IT WORKS
A person on Lucky Ranasinghe’s team at MADITECH reads your brief, finds the right match and replies by email. Sending a request commits you to nothing.
The task, the size and how long you need it. It is fine if you do not know the specs.
We match it to the right setup across established cloud and GPU providers.
You get the configuration, access details and a quote by email. Nothing is provisioned or charged until you accept.
02 / WORKLOADS
For developers, researchers, AI teams, editors, artists, students and companies. Hours or weeks, start with the work you need to do.
Tell us about the model, dataset and memory needs.
Describe the model, input volume and expected output.
Share your software, scene size and deadline.
Explain the pipeline, data size and storage needs.
Tell us which apps you use and how you want to connect.
Describe the task. You do not need to know the hardware.
03 / YOUR BRIEF
You bring the workload. We work through the setup, provider and access details.
Start with the software, the task and your deadline. Leave the technical fields blank if you do not know.
This is a new concierge service. We have no supply listed yet. Every request is reviewed by hand.
We use what you send only to reply to this request. See our privacy policy.
04 / CAPACITY PARTNERS
Tell us what you can offer. We are building relationships with providers and capacity partners, and review each offer by hand.
Hardware details help us assess a possible match. An offer is not a public listing or a promise of paid work.
We use what you send only to reply to this request. See our privacy policy.
Pricing
Every quote names the provider and prices your job three ways. Prices below are for an H100-class GPU and move with the market.
about $1.99 per GPU-hour
The cheapest capacity. The host can stop it at short notice, so your job must save progress.
about $2.99 per GPU-hour
Reliable capacity that stays yours while you pay, with access, a starter environment and a handover note.
about $3.99 per GPU-hour + $149 setup
We build the environment to your spec, watch the run, restart on failure and hand back the outputs.
Remote workstations: $35, $55 or $95 a month. Your quote is fixed for 7 days. Nothing is provisioned or charged until you accept.
No. We match your workload to established cloud and GPU providers. We do not run our own data centre.
It depends on the hardware, hours and data involved; see pricing for the three ways we run a job. Your exact price is in your quote. Nothing is charged by this site, and nothing is provisioned or charged until you accept.
Tell us your privacy and compliance requirements in the brief. The quote names the proposed provider so you can check their terms and decide whether the setup meets your requirements.
Yes, you can request one for a few days. Tell us which software you need, your preferred access method and your dates. We will look for a suitable setup and explain access in the quote.
Yes. Use the provider form to describe your hardware, hosting location and availability. Our team will review it by hand. Submitting does not create a listing or guarantee a match.