Inference Resource Access

Purpose

The company provides AI resources so employees can experiment through approved company systems.

Dev can help employees choose and access the right resources for a prototype. Requesters should focus on the problem they are trying to solve. Dev will help choose the model, tool, or supporting infrastructure that fits the work.

What Dev can provide

The company supports a wide range of inference options, from top-tier premium models to small open-weight models.

Dev can also provide access to supporting AI infrastructure under Dev management and control. Examples include:

• AI-compatible search, for finding relevant information from approved sources;
• fetch tools, for retrieving information from approved sources;
• vector databases, for storing searchable AI-ready information;
• other supporting infrastructure for prototypes.

Some prototypes need a powerful model. Other prototypes work better with a smaller, faster, or more specialized option. Bring the problem to Dev. Dev will help identify the right solution.

Request path

Request company AI resources through the official Dev request form.

Provide enough context for Dev to suggest the right option. If Dev needs more information, they will schedule a follow-up.

A useful prototype request includes:

• what you are trying to build or test;
• what kind of input the AI will receive;
• what kind of output you need;
• any known constraints, such as speed, quality, or context size, meaning how much information the model needs to consider at once;
• whether you already have a model, tool, search, fetch, or vector database need in mind.

Model and resource selection

There is no fixed baseline model tier for all projects.

Different problems need different resources. Dev will help match the prototype to the right option based on the problem, the input and output needs, and any known constraints.

Specialized access

Some requests need additional review because they create more risk or operational impact.

Examples include:

• top-tier premium models;
• high quotas;
• very large context windows, meaning the model needs to consider a large amount of information at once;
• image, audio, video, or multimodal capabilities;
• agentic tool use;
• production-grade API keys or service accounts;
• external vendors outside standard company systems;
• fine-tuning;
• custom model hosting;
• persistent vector storage, meaning searchable AI-ready information that is stored beyond one test session;
• production integrations.

These capabilities are available when justified by the problem. The request should explain why the capability is needed.

Production boundary

Access to AI resources for experimentation is different from approval for production use.

If the AI workflow will access production systems, use sensitive data, or affect customers, employees, money, security, or operations, start the production handoff process before continuing.

Capacity management

Company AI capacity is managed by Dev.

Requesters should focus on the problem they are trying to solve. Dev will choose an option that balances quality, speed, capacity, and safety.

Top-tier model access should go to cases where the extra capability matters for feasibility, quality, safety, or business value.

Access review and expiration

Non-production access may have an expiration or review date.

If the work is abandoned, access should be removed. If the work is moving toward production, access should transition into the production handoff process.