Launching today
Fabric by Carmel Labs
Submit a job via API or dashboard and get results for cheap
32 followers
Submit a job via API or dashboard and get results for cheap
32 followers
Compute platform for AI and data workloads. Submit jobs via SDK or dashboard, get results, no infrastructure management. Run embeddings ($0.0001/text), transcription ($0.001/file), web scraping ($0.001/request), ML training, CI/CD builds, image processing. 80% cheaper than big cloud. No cold starts. Autoretry. Scale whenever. Replace Lambda, SageMaker, Colab, GitHub Actions with one platform. pip install fabric-sdk or use our dashboard. Built for developers who ship, not manage infrastructure.
Hello everyone!
My friend and I built Fabric because we were tired of two things:
Paying AWS $500/month for infrastructure used 10% of the time
Spending more time managing infra than actually building product
The problem we solved:
As a developer, you need to run batch jobs constantly: generate embeddings, transcribe, scrape data, train models, run CI/CD builds.
But to do ANY of this, you first become an infrastructure engineer: spin up instances, manage GPU availability, debug cold starts, babysit long-running jobs, and pray your AWS bill doesn't explode.
We just wanted to run a Python function and get a result. Why was that so complicated?
What Fabric does:
Submit a job via API or dashboard. We run it. You get a result. Done. Since the infra is distributed, the cost of your workload is 80% cheaper than big cloud.
just pip install the sdk client
client = (api_key="your_key")
Generate embeddings for 10,000 documents
result = client.submit_job(
workload_type="embedding_generation",
params={"texts": documents, "model": "minilm"})
No EC2. No SageMaker. No surprises.
More docs on dashboard
How it evolved:
Started as "just embeddings".
Then a friend asked: "Can it transcribe audio?" Added that.
Another: "Can it scrape with real browser fingerprints?" Added that.
Mobile dev friends: "GitHub charges $0.08/min for iOS builds—can you help?" Built CI/CD on real Apple Silicon that is $0.01/min.
Now Fabric runs 10+ workload types through one API: embeddings, transcription, scraping, ML training, CI/CD, image processing, monitoring, data pipelines, custom Python. That is why we give developers workloads that just work.
What makes it different:
Fixed pricing: $0.0001 per embedding, $0.001 per transcription. Know your costs upfront.
No infrastructure: No instances, no capacity planning, no cold starts.
Instant grow: Process 1 item or 100K items, same simple experience.
Auto-retry: Jobs checkpoint and recover on failure automatically.
One API for everything: Replace Lambda, SageMaker, ScraperAPI, GitHub Actions with one platform.
Real problems we're solving:
Startup burning $2K/month on OpenAI embeddings now spends $200 on Fabric.
Mobile team paying $500/month for GitHub Actions iOS builds now pays less than $100 on Fabric's real M1/M2/M3 Macs.
Researcher waiting 3 days in HPC queue now has training runs start immediately on Fabric.
Agency managing proxy pools for scraping now lets Fabric handle it with real residential IPs.
DevOps team debugging Lambda cold starts at 3am now has Fabric jobs start instantly, every time.
Try it now:
https://fabric.carmel.so/
What we'd love your feedback on:
What workloads are eating your time or budget right now?
What's missing that would make Fabric essential for your stack?
Thanks for checking it out!