Rumman Firdos · AI & Cybersecurity Researcher
This section is a structured record of the technology companies, research initiatives, open source organizations, products, community initiatives, and advisory roles I have built, led, advised, or contributed to. It is not a marketing showcase — the emphasis is on the work itself: what each organization or initiative exists to do, the role I played, and how it fits into the broader arc of my work in AI engineering and cybersecurity.
The ventures are grouped by category so the different kinds of involvement are easy to distinguish. Technology companies and products tend to be commercial efforts with a clear customer or deployment context. Research initiatives are oriented toward a specific problem or method. Open source organizations focus on tooling and implementations meant to be reused and audited by others. Advisory and investment roles are listed separately because they reflect a different kind of engagement — input and backing rather than day-to-day ownership.
Several of these ventures are closely tied to the research and writing on this site. Scalewidth, for example, is the vehicle for the autonomous detection infrastructure that appears in the research pipeline, and Exertus is where a lot of the open-source evaluation harnesses and research implementations live. Brevton Group sits at the intersection of AI product work and enterprise software. Where a venture has its own page, that page goes into more depth on the technology, timeline, and current status.
Scalewidth — Founder. Autonomous detection infrastructure for enterprise security operations. The platform treats detection as a software engineering discipline: detection-as-code, CI/CD for rules, and a feedback-driven policy engine that continuously tunes itself from operational feedback. It is built on a streaming event-processing layer (Kafka and ClickHouse) with detectors written in Rust for performance; the policy engine uses online learning to adjust detection thresholds and retire ineffective rules automatically; and a causal attribution layer is intended to distinguish correlation from causation in alert generation. Scalewidth also encompasses the Vigilante security product line and the autonomous SOC platform, both under active development and deployment.
Brevton Group — Founder. A technology company focused on AI engineering and software development. It builds AI-powered products and enterprise software with a focus on helping businesses apply AI in practical ways, spanning LLM-based applications and custom engineering for business clients. The company was founded in 2024, starting as a software development shop and growing into a full tech and AI company. Research on LLM security, prompt injection defense, and adversarial robustness informs how these systems are designed and deployed.
Exertus — Founder. An open-source engineering organization building cybersecurity tooling, AI technologies, developer tools, and research implementations. The focus is on practical, usable tooling for security practitioners — detection engineering frameworks, security automation, and applied AI research code — built on a modern Python and machine learning stack (PyTorch, JAX, Docker, and ONNX Runtime) with a modular plugin architecture. The organization exists to make security engineering and adversarial ML research reproducible and accessible, publishing evaluation harnesses, detection tooling, and research code so the community can use, audit, and build on it.