Cloud-agnostic machine type recommender for computational workloads.
Like a load balancer for cloud instances that recommends the best machine type for your workload across AWS, GCP, and Azure, and stays current as providers deprecate and release new types.
Built with bioinformatics and batch workflow orchestration as the primary use case, but designed to be domain-agnostic.
New here? Try the one-click demo, or read the launch post: Why I built cloudfit for the gap in existing free tooling (Compute Optimizer, Recommender, Advisor) and what cloudfit does about it.
| Package | Description | Status |
|---|---|---|
cloudfit-core |
Scoring engine · workload profiles · hard floor filters · region-aware · fit-based scoring (v0.3+) |  |
cloudfit-provider-gcp |
GCP Compute Engine machine type fetcher (multi-region capable) |  |
cloudfit-provider-aws |
AWS EC2 instance fetcher | planning phase |
cloudfit-api |
REST API. /recommend · /instances · /providers · /diff. Multi-region snapshot bundled. |
 · live demo ↗ |
cloudfit-ui |
One-click Gradio demo over the scoring engine. Workload profile in, ranked instances out. | live demo ↗ |
flowchart TD
GCP["GCP Compute Engine<br/>+ Cloud Billing Catalog"]:::live
AWSAZ["AWS EC2 · Azure VMs"]:::planned
PGCP["cloudfit-provider-gcp<br/>fetch → normalize → price"]:::live
PAWS["cloudfit-provider-aws"]:::planned
SNAP[("MachineType data<br/>bundled JSON snapshot, multi-region<br/>Postgres registry planned")]:::live
CORE["cloudfit-core · scoring engine<br/>hard-floor filter → weighted score<br/>cost · performance · availability"]:::live
API["cloudfit-api · FastAPI service<br/>/recommend · /instances · /providers · /diff"]:::live
UI["cloudfit-ui · Gradio demo<br/>form input · ranked table"]:::live
DOCS["Swagger UI · /docs"]:::live
HTTP["curl / HTTP clients"]:::live
SDK["Python · import cloudfit"]:::live
BROWSER["Browser · one-click demo"]:::live
GCP --> PGCP --> SNAP
AWSAZ -.-> PAWS -.-> SNAP
SNAP --> API
SNAP --> UI
API -->|ranks with| CORE
UI -->|ranks with| CORE
API --> DOCS
API --> HTTP
UI --> BROWSER
CORE --> SDK
classDef live fill:#10261d,stroke:#4dff91,color:#e8eaf0;
classDef planned fill:#191919,stroke:#777,color:#aaa,stroke-dasharray:4 4;
Solid green = shipped · dashed grey = planned.
from cloudfit import WorkloadProfile, MachineType, rank
profile = WorkloadProfile(
vcpu=60,
ram_gb=224,
archetype="io", # io | cpu | mem | gpu | burst
optimize_for="balanced", # cost | performance | availability | balanced
region="us-central1", # optional region hard floor (v0.2+)
)
# Candidate instances come from a cloudfit-provider-* package
# (e.g. `pip install cloudfit-provider-gcp`), or supply your own list:
candidates = [
MachineType(id="t2d-standard-60", provider="gcp", vcpu=60, ram_gb=240, price_hr=2.31, region="us-central1"),
MachineType(id="c3d-standard-60-lssd", provider="gcp", vcpu=60, ram_gb=240, price_hr=3.39, region="us-central1"),
]
best = rank(profile, candidates)[0]
print(f"{best.instance.provider} {best.instance.id} ${best.instance.price_hr}/hr score: {best.score}")
# → gcp t2d-standard-60 $2.31/hr score: 0.9782samplesheet-parser: Format-agnostic Illumina SampleSheet parser (BCLConvert V2 + IEM V1)clinops: Clinical ML data quality library
Author: Chaitanya Krishna Kasaraneni · Google Scholar · ORCID 0000-0001-5792-1095