Note
Activity is a starting point.
A workflow running for longer doesn’t mean it did more useful work. First: what does a good result look like, and how does someone check it?
Notes on measurement →$~:whoami> matheo
Security engineering
for the data and
AI era.
I’m Matheo, a security and AI engineer, previously at Netflix, Anduril, and Okta. I build data pipelines AI can trust and AI workflows people can trust, so only relevant, actionable findings reach a human.
A little curiosity goes a long way. Press ~ for a terminal.A little curiosity goes a long way. Tap >_ for a terminal.
Selected work
Case studies, mostly anonymous: the problem, the stack, the decisions, and the evidence.
Security intelligence & agentic AI
SitRep: AI security-intelligence briefings
Turning a high-volume stream of security news and research into searchable, grounded briefings, where every CVE and figure in a summary has to appear in its source.
Grounded AI & data platforms
Grounded AI for a relocation research tool
An AI mode that answers from ingested official data and deterministic calculators, with a separate fact review before anything generated is published.
AI security research
A lab for attacks on multi-agent AI
A hands-on environment showing how a prompt injection travels through a chain of AI agents with file access, and what different guardrails do and do not stop.
Data quality & measurement
An operational measurement pipeline
Collecting usage and cost without confusing missing data, observed activity, and estimates of value.
Other builds: A real-time collaborative application · A local-first learning platform
What I work on
Where security, data, and AI meet. Each area links to the work behind it.
Grounded AI pipelines
AI that answers from data, not memory: intent routing to database-backed tools, deterministic calculations, and a separate review before anything generated is published.
Agentic AI security
Agent chains for security data, and security for agent chains: allowlisted fetches, claims checked against their sources, and labs that show how prompt injection travels between agents.
Trustworthy data pipelines
Collect, normalize, deduplicate, and quality-check fragmented sources, so missing data stays visible instead of quietly becoming a zero.
Security architecture
Zero trust, identity and access, endpoint security, and cloud migration, from security engineering roles at Netflix, Anduril, and Okta.
How I build
← Back to what I work onAn engineer.
Still curious.
Security is my starting point. I spent years in security engineering at Okta, Anduril, and Netflix, working on zero trust, identity, endpoint, and cloud security. Now I focus on how data, software, and AI fit together, and on making the result trustworthy for the people who rely on it.
World explorer, mentor, and tinkerer. Pushing to prod since Win95. I make room for community work, playful experiments, and learning something outside my usual lane.
Note
Sometimes less software helps.
A small tool can keep preferences on the device, work without an account, and leave less to maintain.
More on the blog ↗Community
Security people can use.
Helping older adults recognize scams, protect themselves online, and know where to turn when something goes wrong.
Community & resources →$ contact --channel linkedin
Working on security, data, or AI?
Let’s compare notes.
© 2026 Matheo Lord-Martinez
$~:ls playground/
Arcade 95
Games, desk accessories, and a terminal with a few secrets, inside a little computer from a different era.
- Pinball
- Snake
- Tetris
- Breakout
- Asteroids