AI/ML Infrastructure Build

WANAC Foundation is building a private AI platform to support transitioning service members and veterans. We have a dedicated on-prem GPU server and we want a skilled volunteer to set it up “as if new” and bring it to full production potential.

We want to create a secure, repeatable, production-grade AI stack that includes:
1. Core infrastructure and GPU readiness

• Fresh OS baseline and GPU driver/CUDA stack validation
• Container runtime and GPU scheduling setup (Docker and/or Kubernetes/K3s)
• GPU monitoring and health checks (utilization, temps, failures, alerts)

2. Model serving (inference)

• Production LLM serving for multiple models/services (OpenAI-compatible endpoint preferred)
• Ability to run multiple workloads concurrently (general assistant, program assistant, embeddings/RAG)
• Simple routing and resource allocation plan across GPUs

3. Retrieval and knowledge services (RAG)

• Embeddings service + vector database integration
• Document ingestion pipeline for WANAC policies, SOPs, program content, and templates
• Grounded responses with citations and access control

4. Fine-tuning pipeline (LoRA/QLoRA)

• Repeatable training workflow (data prep → train → evaluate → deploy)
• Recommended model sizes/configs appropriate for GPU constraints
• Basic evaluation harness and deployment procedure for fine-tuned models

5. Security, access control, and operations

• Role-based access approach aligned to our identity stack
• Logging/auditability for service requests and model changes
• Backup/restore strategy for configs and model artifacts
• Runbooks: deploy/upgrade/rollback/troubleshoot

Deliverables at the end of the engagement
• A working production-style deployment (MVP) that WANAC can use immediately
• Architecture diagram and documented “day-2 operations” runbook
• A prioritized 90-day roadmap to expand capabilities after the initial build

WANAC serves transitioning service members and veterans. A private, production-grade AI server enables us to:
• Provide consistent, scalable support through AI-assisted coaching and structured pathways.
• Protect privacy by keeping sensitive data under WANAC control rather than relying on third-party AI platforms.
• Automate follow-up and milestone tracking so participants do not fall through the cracks.
• Produce credible outcome reporting for schools, employers, funders, and partners.
• Reduce long-term costs by running core AI services internally.

This build will become foundational infrastructure for delivering veteran programs with discipline, quality, and measurable impact.

How will you set this project up for success?
• We will treat this as a priority infrastructure project with a dedicated point of contact and fast decision-making.
• We will provide full server access, hardware details, and a clean-build starting point (“assume new server”).
• We will hold a structured kickoff to confirm scope, timeline (4–6 weeks), and milestones.
• We will meet weekly and respond quickly to unblock decisions.
• We will provide a shared project folder and a single task tracker to keep work organized.
• Data discipline: we will not share sensitive participant data during setup; we will use sanitized documents for testing.