2,000 engineers ran on costly commercial AI with no control over data paths. Self-hosted GLM on dedicated GPUs cut AI costs 90%.
A digital bank supporting around 2,000 engineers who relied on commercial AI services from OpenAI and Anthropic for day-to-day development work.
Running engineering workloads on commercial AI services came at a high cost and left the company without control over the models or the data path...
Also reported by
Sources found with NewsCatcher's Web Search API
Digital banking and financial services app for payments, savings, and investments.
Provided the GLM 5.3 open-weight model that Revolut runs on its own GPU infrastructure for coding tasks
Revolut's Software development is part of this use case:
Related implementations across industries and use cases
Manual processing blocked 24/7 support. Now, an AI agent resolves routine requests autonomously, freeing staff for relationship work.
Each integration task took 2–3 hours, crowding out innovation on a 650 TPS platform. AI agents now handle everything from design to production in minutes.
Engineers capped at basic chat, complex work out of reach. Now 700 tackle cross-jurisdiction compliance reviews in hours, not weeks.
Each integration task took 2–3 hours, crowding out innovation on a 650 TPS platform. AI agents now handle everything from design to production in minutes.
Deciphering legacy systems bottlenecked senior reviewers. Now, early-career devs use Claude to navigate code, freeing seniors to build.
Buyers discovered missing approval steps midway and restarted from scratch. A single intake door with real-time AI correction ended that.
Most AI tools addressed one step—code review or a single feature—and couldn't run in-house. Factory covers the full SDLC, on-premise.
Tournaments running simultaneously meant an hour of manual checks each. AI agents now run them in minutes, freeing the team to be proactive.
Large AI training jobs meant fighting for preemptible slots or leaving campus. Marlowe gave any lab guaranteed multi-node access on demand.
2,000 engineers ran on costly commercial AI with no control over data paths. Self-hosted GLM on dedicated GPUs cut AI costs 90%.
A digital bank supporting around 2,000 engineers who relied on commercial AI services from OpenAI and Anthropic for day-to-day development work.
Running engineering workloads on commercial AI services came at a high cost and left the company without control over the models or the data path...
Also reported by
Sources found with NewsCatcher's Web Search API
Digital banking and financial services app for payments, savings, and investments.
Provided the GLM 5.3 open-weight model that Revolut runs on its own GPU infrastructure for coding tasks
Revolut's Software development is part of this use case:
Related implementations across industries and use cases
Manual processing blocked 24/7 support. Now, an AI agent resolves routine requests autonomously, freeing staff for relationship work.
Each integration task took 2–3 hours, crowding out innovation on a 650 TPS platform. AI agents now handle everything from design to production in minutes.
Engineers capped at basic chat, complex work out of reach. Now 700 tackle cross-jurisdiction compliance reviews in hours, not weeks.
Each integration task took 2–3 hours, crowding out innovation on a 650 TPS platform. AI agents now handle everything from design to production in minutes.
Deciphering legacy systems bottlenecked senior reviewers. Now, early-career devs use Claude to navigate code, freeing seniors to build.
Buyers discovered missing approval steps midway and restarted from scratch. A single intake door with real-time AI correction ended that.
Most AI tools addressed one step—code review or a single feature—and couldn't run in-house. Factory covers the full SDLC, on-premise.
Tournaments running simultaneously meant an hour of manual checks each. AI agents now run them in minutes, freeing the team to be proactive.
Large AI training jobs meant fighting for preemptible slots or leaving campus. Marlowe gave any lab guaranteed multi-node access on demand.