I don't just ship projects.
I document decisions, trade-offs, and outcomes.

My digital garden built around case studies, not screenshots.

AI Scientist and AML Researcher · Home Profile

Explore Case Studies How I Make Decisions
Case Study · Principal Adversarial AI Architect & Red Team Lead Project M.A.L.I.C.E M.A.L.I.C.E. (Malicious Agentic Linkage & Infrastructure Corruption Engine) is an advanced, automated offensive framework designed to exploit multi-agent orchestration ecosystems and local AI infrastructure (such as Ollama and LangChain), exposing catastrophic vulnerabilities in localized agentic workflows. Case Study · Lead AI Architect Project A.L.I.C.E A.L.I.C.E. (Autonomous Linked Intelligent Compute Engine) is a simple, hands-free, 100% locally running voice assistant that enables natural spoken interaction with Ollama LLMs, designed with future secure remote access in mind.

Founder, AI Scientist, Red Team Lead and AML Researcher @NoctisResearch. Focused on secure Cloud-Native AI Systems Architecture and the convergence of artificial intelligence and humans.

Every major decision is documented, because great researchers don't just build things, they explain why.

Specialization: Adversarial Machine Learning (AML) Infrastructure & Agentic Ecosystem Exploitation.

Read my journey
W Mouton W Mouton – anime

Recent Decisions

View all
May 2026

The ZERO-TRUST LOOP: Sandboxing Multi-Agent Tool Execution

Traditional application security relies on input sanitization at the boundary. However, in an agentic ecosystem, the threat is inverted: the LLM is the boundary, and the data it consumes from external tools (web scrapers, database connectors, file parsers) acts as an unverified execution vector. If an agent processes an Indirect Prompt Injection (IPI) hidden in a public web page, it can be manipulated into executing malicious host-level commands via its own file-system or shell tools. We require an infrastructure model where agent tool execution is decoupled from the host and strictly containerized.

Mar 2026

The GHOST-NODE Protocol: Non-Attributable Infrastructure

Traditional 'Secure Servers' are stationary targets; they have IP footprints, physical locations, and detectable power signatures. In high-stakes defense and biological data management, a compromised server is a terminal failure. We require a 'Dark-Infrastructure' model where the compute environment has no fixed topological position and utilizes 'Ambient Processing' to operate without leaving a detectable trace.

What People Say

"Will is an exceptional engineer who pairs technical brilliance with long-term vision. He doesn’t just solve problems; he builds a knowledge base around them. By documenting the 'why' behind every solution, he’s empowered the team to scale efficiently and effectively eliminated the cycle of repeating past mistakes."

Snoweirdo, Sr. Design Engineer at Plank DesignsCo-founder, 2024-Present

"Working with Will has been a career-defining experience. He didn't just help me meet my goals—he helped me shatter them in half the expected time. His ability to implement innovative scaling strategies completely transformed how I approach my business."

Problems I Enjoy Solving

Scaling systems without scaling complexity · Navigating consistency vs availability · Modernizing legacy models & codebases · Designing durable APIs for ML · Building teams that ship · Turning technical debt into strategic investment · Documenting the "why" for future-you

Let's Work Together

If you need someone to bridge the gap between complex AI research and scalable production architecture, let's talk.

View My Work Get in Touch