> lab
RW Cognition Labs
Engineering intelligent systems with clarity, discipline, and scientific rigor. We work on agentic architectures, cognitive tooling, and applied research around controllable AI.
/ applied research · agentic systems · cognitive architectures
> capabilities
[01] cognitive
Cognitive Engineering
Designing intelligent systems with explicit reasoning, memory, and control surfaces instead of opaque, single-shot pipelines.
[02] research
Applied AI Research
Investigations, experiments, and ablations around agent behavior, failure modes, and control mechanisms in real environments.
[03] systems
Systems Implementation
Building infrastructure, tooling, and scaffolding that make complex intelligent systems observable, testable, and maintainable.
> principles
We treat models as computational artifacts, not oracles.
We care about controllability, observability, and reproducibility more than demos.
We avoid anthropomorphism and hype; we work from mechanisms and constraints.
We build systems that are understandable, debuggable, and correctable by engineers.
> active projects
Autonomy Bench
A controlled environment for probing agent behavior, failure modes, and controllability under varied tasks, tools, and constraints.
focus: evaluation · safety envelopes · failure taxonomies
Context Graph Engine
A semantic control-surface map for understanding memory, state, and influence channels across agent workflows and tools.
focus: state tracking · influence mapping · control surfaces
Agentic Infrastructure Toolkit
Tools and patterns for building agent workflows, capability registries, and deterministic scaffolding for complex reasoning loops.
focus: scaffolding · registries · deterministic flows
> lab notes
- [2025-02-04] Notes on multi-step toolchains and hidden state draft
- [2025-01-18] Failure modes of naive agent retry loops internal
- [2025-01-12] Short notes on cognitive architecture layering internal
- [2024-12-27] Minimal deterministic agent scaffolding internal
>> Some notes will be published as public write-ups. For now, this is a glimpse of the working log.
> founding team
RW Cognition Labs was founded by Roger Bush and Wael Ghandour, engineers shaped by the practical realities of Silicon Valley: systems that break under scale, architectures that only survive with discipline, and environments where correctness and failure modes are not theoretical concerns but daily constraints. The lab’s foundation comes from decades spent designing, securing, and repairing real systems—not talking about them.
Roger and Wael come from different domains—large-scale distributed systems and cognitive architectures on one side; security engineering, adversarial thinking, and intelligent control surfaces on the other. Their overlap is why the lab exists: both have spent their careers building and defending the kinds of systems that demand depth, clarity, and a refusal to hand-wave complexity.
Modern AI systems sit at the intersection of distributed computation, cognitive modeling, and adversarial pressure. Most teams only bring one angle. The founding team brings all of them, informed by years inside the Valley’s highest-velocity environments and hardened by the operational constraints those systems impose.
two backgrounds, one principle: build systems that stay intelligent under pressure.
> connect
If you are working on serious problems in intelligent systems, autonomy, or cognitive tooling and want engineering-level collaboration rather than generic consulting, reach out.
email: [email protected]