Patryk
CEO
Built the GPU network deAPI runs on: 820,000+ accounts since 2017, and the move that turned it into an AI inference layer. At deAPI he decides what gets built and what it costs.
in/patryk-puschAbout Us
One API. A distributed GPU network. Run image, video, audio, and text models without managing infrastructure, and at a fraction of traditional cloud cost.
Mission
AI compute should be priced like a utility, not a luxury.
A handful of cloud providers own the GPUs that modern AI runs on — and they set the price. We took a different route: we route inference through a global network of idle consumer GPUs. The result is the same open-source models, with no GPUs to provision and no capacity to request, for up to 20× less.
The problem
The bottleneck isn’t models. It’s GPUs.
Open-source models are now competitive with closed ones. Running them isn’t.
GPU capacity is expensive, concentrated in a few hyperscalers, and wrapped in lock-in.
Most teams either overpay or never ship.
Meanwhile, hundreds of thousands of high-end GPUs sit idle in gaming PCs every night.
The solution
One API. A distributed GPU network.
deAPI aggregates idle consumer GPUs into a single inference layer and exposes it through one consistent API. We handle GPU orchestration, execution, and scaling — you call an endpoint. Built on a GPU network live since 2017, it turns underused gaming hardware into production compute.
Already on the OpenAI SDK? deAPI ships an OpenAI-compatible endpoint at oai.deapi.ai/v1 — change base_url and api_key, keep your code.
Team
Sławomir
CTO
Built and ran the engineering organization behind SocialTalent for 13 years — used by IBM, Siemens, Cisco and Nokia across 60 countries — and founded codelabs.rocks, a 30-person engineering firm. At deAPI he owns the architecture of the inference layer.
in/solszynski
Alex
COO
A decade scaling fintech and distributed infrastructure; founder of AngelBlock. Based in the SF Bay Area, he runs deAPI’s operations and investor relationships.
in/alexstrzesniewski
Dawid
Software Developer
Built the desktop client for the network deAPI runs on. At deAPI he owns the worker — the software that turns a gaming PC into an inference node.
Adam
Senior Developer
Architected the system that keeps thousands of nodes connected into one distributed compute network. At deAPI he decides which open-source models make the catalog — evaluation, quantization, and the parameter tuning that determines what each GPU can actually run.
Patryk A.
Back-end Developer
Former CTO of a CPA affiliate network with 850,000+ publishers across 154 countries. At deAPI he built the client dashboard from scratch and owns backend performance and security.