# Software built to run for decades.

Kushawaha Group builds and keeps infrastructure software: GPU schedulers, governed AI brokers, agentless telemetry and zero-trust gateways.

[Talk to the systems team](https://kushawaha.com/contact)

Try it: lower the ceiling, then send until the broker refuses.

research-agent

Inference broker · October

Streaming

** $8.42** spent of **$10.00** monthly ceiling

Ceiling $10

1. 09:41:07 POST /v1/messages 200 28,670 tok · $0.43
2. 09:39:52 POST /v1/messages 200 19,330 tok · $0.29

## We build the layers underneath.

Your models, products and teams run on top. Each layer below is software the group writes, ships and keeps running.

Your workloads run here

1. AI that stays on budget Signed requests, hard monthly ceilings, HTTP 429 when spent
2. Zero-trust gateways Client identity headers stripped at the edge
3. Agentless telemetry One-second SSH probes, zero daemons on the host
4. GPU schedulers CUDA, ROCm, Metal and Arc in one gossip cluster
- Five operating units
- [Inference brokering](https://kushawaha.com/capabilities/autonomous-ai) · AI governance · Ed25519, HMAC, spend caps
- [GPU scheduling](https://kushawaha.com/capabilities/distributed-compute) · Distributed compute · Gossip, single binary, no CGO
- [Agentless telemetry](https://kushawaha.com/capabilities/autonomous-sre) · Observability · SSH, sysfs, NVML, ROCm
- [Zero-trust gateways](https://kushawaha.com/capabilities/zero-trust-security) · Security · Header sanitising, signed tokens
- [Workforce platforms](https://kushawaha.com/capabilities/enterprise-platforms) · Enterprise software · Offline-first PWA, IndexedDB
- NVIDIA CUDA
- AMD ROCm
- Apple Metal
- Intel Arc
- Gossip discovery
- Single binary
- Zero CGO
- Ed25519 tokens
- HMAC attestation
- HTTP 429 ceilings
- sysfs probes
- NVML
- Zero agents
- Offline-first PWAs

## Inside the systems

How each system works in production: what it runs on and what happens to a request.

Distributed systems

### GPU scheduler

Under 15 ms median job handoff on a local network.

Hardware

NVIDIA CUDA, AMD ROCm, Apple Metal, Intel Arc

Coordination

Encrypted peer-to-peer gossip, no master node

Footprint

21.4 MB single binary

Handoff

< 15 ms median

1. Node starts the binary and finds peers over UDP gossip
2. Driver probe reads VRAM and temperature from sysfs, NVML or Metal
3. Queue moves the job to the lowest-latency free GPU
4. Job runs in an isolated process sandbox

## Owned for the long run.

### How we hold

The group keeps every system it builds. The team that ships a scheduler or a gateway stays to run it, patch it and answer for it, with no exit plan and no rewrite every two years.

### How we engineer

Single binaries, memory-safe languages, signed requests between services and hard ceilings on anything that can spend money. Each rule is written down as an RFC before it ships.

[Read the operating model](https://kushawaha.com/about)

## Research and writing

Every rule we ship is written down first: open RFCs and notes from the engineers who run the systems.

[All RFCs](https://kushawaha.com/research) [All articles](https://kushawaha.com/blog)

[RFC · October 2026 RFC-2401: In-App LLM Proxy Architecture with Enforced Dollar Ceilings This specification outlines an in-app proxy protocol for running untrusted autonomous agents without sharing provider API keys. By validating requests through an internal broker that checks signed Ed25519 tokens and enforces hard monthly spend caps, systems prevent key theft and unbudgeted inference charges. Read the RFC](https://kushawaha.com/research)

1. [Article · October 18, 2026 Eliminating Inference Key Leakage with In-App Cryptographic Brokers](https://kushawaha.com/blog/eliminating-inference-key-leakage-with-in-app-brokers)
2. [RFC · August 2026 RFC-2388: Decentralized Gossip Scheduling Across Mixed GPU Architectures](https://kushawaha.com/research)
3. [Article · September 24, 2026 Scheduling Heterogeneous GPU Clusters Without Central Masters](https://kushawaha.com/blog/scheduling-heterogeneous-gpu-clusters-without-central-masters)
4. [RFC · July 2026 RFC-2340: Agentless Server Telemetry Using Ephemeral Compound Queries](https://kushawaha.com/research)

---

Source: https://kushawaha.com/
Published by Kushawaha Group. Index of all pages: https://kushawaha.com/llms.txt