Welcome to Spare Cores!

At Spare Cores, we believe cloud infrastructure shouldn't feel like a guessing game—especially for data science and machine learning teams. Our mission is simple: help you run every ML workflow faster, cheaper, and with less hassle, by turning real usage data into smart, actionable decisions.

We're building tools that give teams clarity, control, and cost-efficiency—without the overhead of traditional DevOps. And we're doing it the way we wish every infrastructure tool worked: transparent, developer-first, and workflow-native.

Why we started

Our founder, Gergely Daróczi, spent more than a decade managing infrastructure for data science teams. In that time, he saw the same problems again and again: engineers over-provisioning cloud instances "just to be safe", expensive GPUs sitting idle, and workloads delayed by ticket queues and trial-and-error. The costs added up quickly—but even more painful was the time lost to debugging cloud infrastructure configs instead of building models.

Spare Cores started with a simple idea: what if teams could make smart cloud decisions automatically, based on what their jobs actually use—not what they guess they'll need?

We began by building a public benchmark engine, measuring thousands of cloud instance types across AWS, GCP, Azure, and Hetzner. That data became the foundation for Spare Cores Navigator, our open platform for exploring real-world price and performance. From there, we launched the open-source Resource Tracker package—a lightweight Python/R integration for Metaflow (with Flyte, Airflow, and other orchestration tools coming soon), that captures real CPU, RAM, and GPU usage, and recommends better-fitting instances for next time.

Today, over 5,000 Metaflow steps per week are tracked with Spare Cores. And we're just getting started.

What we offer

Spare Cores is building a full suite of tools and services to help DS/ML teams cut costs and speed up experimentation—without changing how they work.

Our open-source packages plug directly into your workflows, collecting fine-grained usage data and showing exactly where you’re wasting compute. For teams who want more control and visibility, we offer a cloud-based platform called Sentinel, with two flavors:

• Sentinel Access is our free tier. It gives individual users a dashboard to track their runs, see seven days of history, and get real-time alerts about over-provisioned jobs.
• Sentinel Pro is built for teams. It adds team-wide dashboards, historical reports, cost savings deltas, policy-based automation, and cross-cloud optimization—plus support for SSO and RBAC when you need governance.

Alongside these, we also offer scalable Shiny and Streamlit hosting (on your cloud or ours), cloud optimization consulting, and sizing services for ad-hoc AI workloads like LLM labeling or document processing.

How we're different

There's no shortage of tools that show you where your cloud spend is going. But most of them stop at visibility—and none of them speak the language of ML workflows. Spare Cores lives inside your workflow. No portals to log into. No magic boxes. Just one line of code and actionable results.

We're open by design—our benchmark engine is licensed under the Mozilla Public License (MPL 2.0), the related datasets under Creative Commons Attribution-ShareAlike (CC-BY-SA), and we're committed to transparency, verifiability, and community trust. Our tools don't lock you in. They help you make smarter decisions, faster.

And unlike traditional FinOps tools built for 24/7 SaaS traffic, we're laser-focused on batch-based experimentation.

We aim to be your trusted sidekick—technical, transparent, and always on your side when it comes to cloud costs.

Who we are

Spare Cores is proudly bootstrapped and independent, with early support from a €150,000 EU NGI Search grant. We're a member of the NVIDIA Inception program and have backing from AWS, GCP, Azure, Hetzner, and UpCloud in the form of compute credits and cloud support.

We've been featured in The Pragmatic Engineer, presented at PyData, Berlin Buzzwords, and useR!, and have been lucky to receive early feedback and encouragement from teams at Outerbounds, Too Good To Go, The Duckbill Group, and the European Commission.

Our team brings deep experience in data science, MLOps, cloud architecture, product, and growth. But more than anything, we care about building tools we'd want to use ourselves.

Let's talk

If you're struggling with cloud cost control, buried in infrastructure tickets, or just want to spend less time configuring and more time building—Spare Cores is for you.

Start by installing the open-source Resource Tracker. Sign up for early access to Sentinel Access to see your own 7-day job history. Or grab time with us for a quick 30-minute call —we'd love to hear your pain points and preview what's coming next.

We're here to help you stop guessing, start optimizing, and make every core count.