Work & Partnerships

We are a small, research-first team.
Maaza models and MCPBodega are Apache 2.0 and will stay that way.

We accept 3–5 engagements per year when all three conditions are met:

Edge AI, autonomous systems, or innovative tooling—SaaS or beyond. We look for technical depth, not domain.

Your use case benefits from models under 20M parameters. If frontier models solve it, you don't need us.

Advancing useful, offline-capable agents or contributing to edge AI research. We're research-first.

Typical scope

We build private orchestrators under 20M parameters that run entirely on-orbit—no cloud dependencies, no export issues, and they fit in under 50 MB.

→ Custom NLM/MLM training for your exact toolset
→ Private high-accuracy orchestrator (<20M parameters) with your proprietary MCPs
→ Quantization + deployment hardening (4-bit, 3-bit, edge TPU, MCU)

Rate: Six figures per engagement (fixed-price, scoped in days).

Availability (Q1 2026): 2 slots open.

If you believe your use case matches the above, send a short email to

Subject line format:
[Work] <Your Company – One-sentence summary>

We reply only to requests that meet the criteria.

Thank you.