Singapore energy client: AI compute from hardware to model to helpdesk

NDA-bound Singapore energy client. TWO TWO delivered GPU compute deployment, a client-specific scheduling platform, and integrated GLM + QWEN models into the client's business systems, backed by 24×7 managed operations. 8 compute nodes online — client team focuses on models, not hardware.

Client: Singapore energy client (NDA) Period: GPU deployment + scheduling + model integration + managed ops Services: AI Compute Lease (B300 / RTX PRO 6000), Datacenter Integration & Ops

Context

The client is a Singapore-based energy company that wants to bring AI into its everyday work — natural-language equipment inspection Q&A, automated report generation, intelligent retrieval on engineering knowledge bases. The team knows model selection and how to wire AI into the business. What they don’t want to do: procure GPUs, rack them, run the data centre, write a scheduling platform, or take the night shift.

They needed the whole programme handled: GPUs run for us, model integration into business systems delivered, 24×7 someone on watch.

What we did

TWO TWO owns the full lifecycle — from hardware through model integration to daily operations:

1. Compute deployment

8 compute nodes online — GPU model and capacity sized to the client’s business profile. Rack move-in, power, network, monitoring and spares are all handled by TWO TWO.

2. Compute scheduling

A scheduling platform customised for the client:

Resource isolation between training and inference workloads
Priority / queue / preemption policies tiered by business role
Usage, cost and SLA dashboards for visibility

3. Model integration with business systems

Two production models landed inside the client’s business:

GLM — bilingual (Chinese / English) general capability, used for engineering report drafting and internal knowledge-base Q&A
QWEN — long-context and multi-turn dialogue, used for intelligent assistants inside business systems

Both are exposed to the client’s existing tickets, knowledge base and reporting via API. Business users call AI from tools they already use.

4. 24×7 managed operations

TWO TWO’s on-call team owns node health monitoring, model service availability, performance sweeps and incident response. The client team focuses on model iteration and business integration — not on hardware and ops.

Results

Go-live from “quarters” to “weeks” — hardware, scheduling and model integration under one delivery. The client team is on business from day one
Cost controlled — 8 nodes sized to actual demand, avoiding the “buy a batch first, figure out later” waste
Model integration is not siloed — GLM / QWEN reach existing tickets, knowledge base and reports via API. Business users don’t learn a new tool
24×7 coverage — client team can go home. They’re no longer woken up by GPU alerts

The full case is NDA-bound. Reach out and we’ll walk you through it in private.

Start with the problem. The solution comes after.

A 30-minute call with a solution architect. No spec needed — just tell us where the business is stuck. Systems that don't talk. Data that won't behave. An overseas office with no one running IT. A DR plan on paper that no one has tested. A first AI use case you want to try for real. If we fit, we'll show you the next step. If not, we'll tell you straight.