Product leader, founder, and hands-on builder working where AI/ML meets real-world systems at scale - currently focused on cloud security and multi-agent orchestration.
|
LLM-first cloud log anomaly detection Cloud anomaly detection for GCP Audit Logs and AWS CloudTrail, built for small security teams. Visit Flare → · Evaluation case study → · Original X thread → |
Cost firewall for AI agents An open-source scheduler, budget circuit breaker, and cost-attribution layer for teams running agentic AI in production. Landing page → · Docs → · GitHub → · PyPI → |
|
Security skills for AI coding agents Drop-in skill verification, cost tracking, tool authorization, behavioral anomaly detection, and pre-execution safety checks. |
A private place for what matters Save links, text, and screenshots in one calm library, then bring back the useful context, next move, or brief when it matters. |
Field notes from a month of using two AI coding agents together: the quiet bugs a second model caught, why model disagreement is useful, the metric I started tracking, and where cross-model workflows still fall short.
How I use golden fixtures, source grounding, failure analysis, and regression testing to evaluate an LLM-first cloud-security product.
A 20-seed simulation showing how one scheduling knob trades draining the deepest agent backlogs against fairness, including the point where a third of the agents starve.
Hands-on stress tests of raw Google ADK across runaway loops, growing context, concurrent agents, shared state, and the production controls a managed platform still needs to provide.
How four Flare GitHub Actions extend cloud security from pre-merge checks through deployment verification, continuous monitoring, and incident investigation.
- Tab Review Marker - A free, local-first Chrome extension that adds To Review, Reviewed, and Needs Attention states directly to tab favicons, with notes and AI-ready exports.
- OpenClaw Daily Spending Allowances - Proposed a consumer-friendly, Screen Time-style spending allowance for unattended AI agents, with plugin-owned policy and a privacy-conscious cost-and-provenance contract for OpenClaw core. Awaiting maintainer product review.
- Generate Runbook Skill - A portable Codex and Claude Code skill for creating safe, validated operational runbooks.
- ADK Learning Guide - An interactive deep-onboarding guide to Google's Agent Development Kit, including a LOCO-Agent integration module.
- Meta - Led the messaging developer platform, a billion-dollar-plus revenue opportunity spanning third-party APIs and Ads Manager. Built and shipped an LLM-augmented decision system and quality measurement at scale.
- PlanetWatchers - Founded a geospatial AI startup for monitoring natural resources at scale: deep learning research → production platform → revenue. Planet.com coverage →
- Sumo Logic & AppDynamics - Built ML-powered observability products.
- AI-first threat detection that replaces rules-based SIEMs
- The economics and orchestration of multi-agent systems
- Turning frontier research into safe, interpretable production AI
- Discovering Anomalies and Root Causes in Applications via Relevant Fields Analysis
- A Message-Passing and Load-Sensitive Contention-Based MAC Protocol for Ad-Hoc Wireless Networks