Customer story
Axine Labs runs zero-drift compliance AI on RLC Pro AI
Axine Labs built Lisaris, an agentic platform that turns satellite imagery into legally defensible land-use compliance dossiers, and runs it unquantized on dual-H100 bare metal with RLC Pro AI.
2x H100
Bare-metal GPU node
A dedicated node with two NVIDIA H100 GPUs
70B
Parameters, unquantized
Served with zero tolerance for drift
$6/hr
Per GPU node
Compute cost no longer lost to driver debugging
4
Regulated markets
EUDR, agriculture, finance, and forestry
The platform, the stack, and why it runs on RLC Pro AI
How a compliance AI company serves a 70B model unquantized on bare metal without losing engineering time to the driver stack.
- Who
- Axine Labs, builder of Lisaris, an agentic platform for auditable land-use compliance verification
- What they run
- A two-node bare-metal setup: dual H100s serving an unquantized Llama 70B model with vLLM, plus a dedicated CPU node for PostGIS and geospatial processing
- What’s on it
- RLC Pro AI, running a validated CUDA and driver stack for both GPU inference and training
- What it supports
- Legally defensible compliance dossiers for EUDR, agriculture, finance, and forestry customers across five global regions
- Why RLC Pro AI
- A validated, working driver and CUDA environment from the start, with a security posture that clears enterprise and marketplace audits
“RLC Pro AI gave us a validated, hardened OS with working drivers and CUDA environments immediately. Eliminating the dependency doom loop on day one meant we stopped burning engineering cycles and compute dollars on driver debugging instead of customer deliverables.”
The story in brief
Axine Laboratories Inc. serves the model behind Lisaris Earth Observation Studio unquantized, because regulatory compliance dossiers can’t tolerate drift or hallucination. That raised the stakes on getting the driver and CUDA stack exactly right. RLC Pro AI arrived validated and working, ending the driver conflicts that had caused recurring server panics on standard Linux setups.
Why this matters beyond one platform
The same dependency problems that threatened Lisaris threaten any team running production AI on GPUs.
The dependency doom loop, solved before it started.
A validated driver and CUDA combination that works out of the box beats days of chasing library conflicts.
A custom OS, without giving up bare metal.
Running RLC Pro AI on bare metal replaced a serverless platform that couldn’t run a custom OS at all.
Compliance-grade security, out of the box.
For Axine, the base OS cleared enterprise and marketplace security audits without a separate hardening project.
Evaluate RLC Pro AI for your environment
RLC Pro AI overview
Enterprise Linux tuned and validated for production AI workloads, with the NVIDIA CUDA and DOCA-OFED stack and AI framework configurations tested together before they ship.
RLC Pro AI solution brief
What ships in RLC Pro AI and how teams run GPU inference and training on it.
Talk to CIQ about RLC Pro AI
Tell us about your GPUs and the models you serve, and we’ll show you what running them on a validated driver and CUDA stack looks like in your environment.