Work

Context

Casting is fragmented, inefficient, and heavily manual — especially for indie filmmakers and emerging talent. Actors struggle to understand what they're right for, while productions spend time filtering noise instead of finding fit.

Problem

Traditional casting platforms rely on static profiles and keyword searches, which don't reflect real-world fit, availability, or context. Both sides lose time, and good matches are missed.

System Design

I designed CastIQ as a data-driven casting platform that structures talent profiles, project requirements, and availability into a system that can be reasoned about — not just searched.

AI's Role

• Analyze role requirements beyond simple keywords
• Help talent understand alignment with roles
• Reduce friction in discovery without replacing human decision-making

Why It Matters

CastIQ isn't about automating casting decisions — it's about making better conversations possible by improving signal on both sides.

Context

Public safety dive teams operate in high-risk environments where readiness, training, and compliance aren't optional — but the systems supporting them are often outdated or fragmented.

Problem

Tracking certifications, training dives, readiness, and compliance is often manual, inconsistent, and reactive. That creates risk where there shouldn't be any.

System Design

ORCA | ONE was built as an operational system that centralizes diver readiness, training history, and compliance status into a single, authoritative source of truth.

AI's Role

• Flagging gaps in training or certification
• Identifying patterns in readiness over time
• Supporting leadership with clearer situational awareness

Why It Matters

This system supports real teams doing real work, where clarity and preparation directly impact safety.

Context

AI is moving faster than most organizations can meaningfully track. The conversation is often polarized — either hype-driven or fear-driven — with little grounded analysis.

Problem

There's no simple way to monitor how AI is affecting roles, workflows, and systems over time without drowning in noise.

System Design

AI Watchtower was built as a signal aggregation and analysis platform designed to surface trends, impacts, and early warnings — not predictions.

AI's Role

• Aggregate and summarize signals across domains
• Detect emerging patterns
• Support human interpretation rather than replace it

Why It Matters

AI Watchtower exists to support informed decisions, not hot takes — especially as AI continues to reshape work and systems.

Context

Homeschooling families often lack access to practical, experience-based learning — trades, outdoor skills, arts, farming — that traditional classrooms don't offer. Meanwhile, skilled community members are willing to share their knowledge but have no easy way to connect with families nearby.

Problem

There's no simple, trust-first directory for families to discover local mentors offering in-person, hands-on learning experiences. Existing platforms are either too complex (scheduling, payments, reviews) or focused on online tutoring, missing the human, community-driven connection entirely.

System Design

A public mentor directory with filtering by skill, location, age range, and format. Mentor profiles with inquiry submission for parents. A Skill Request Board where families post learning wishes and others express interest. Manual admin approval for all mentor listings — trust over speed. Built with React, Tailwind CSS, and Lovable Cloud for the backend (database, auth, storage, edge functions).

AI's Role

• Generates the social media preview (OG) image for the platform
• Powers the development workflow (code generation, iteration, deployment via Lovable)
• Intentionally NOT used for matching, recommendations, or content moderation in V1 — human oversight is a core principle

Why It Matters

Children learn best through real experiences with real people. Neighbor Learning puts community trust and human connection first — no algorithms, no ratings, no payments. It's a return to how learning used to happen: from the people next door.

Context

Modern society is critically dependent on digital knowledge, which is predominantly cloud-based. This creates a severe vulnerability: in scenarios like grid collapse, natural disasters, EMPs, or conflicts, essential information becomes inaccessible precisely when it's most needed. Last Light AI was built to address this, providing a resilient, offline repository of human survival knowledge.

Problem

The primary problem is the fragility of digital knowledge in grid-down scenarios. It mitigates the single point of failure inherent in cloud-dependent information systems, ensuring access to vital data even when Google, YouTube, and emergency services are unavailable. Unlike static digital libraries, it offers an interactive AI that can reason through specific survival queries, providing actionable advice rather than just raw data.

System Design

Frontend built with React, TypeScript, and Vite for high-performance development. Styled with Tailwind CSS for responsive, utility-first design. Hosted on Vercel for global availability. A clean, component-based architecture designed for speed and future extensibility, focusing on a "Local-First" philosophy for eventual offline integration.

AI's Role

• Powers the entire development lifecycle, including code generation, rapid iteration, and automated deployment via Lovable
• Generates high-quality visual assets, including OG images and thematic illustrations
• Assists in structuring and indexing the 512GB survival knowledge base for efficient retrieval by local models
• Intentionally NOT used for real-time matching or content moderation — human oversight is a core principle of trust and reliability

Why It Matters

When infrastructure fails, knowledge shouldn't fail with it. Last Light AI ensures that critical survival intelligence remains accessible, interactive, and useful — regardless of connectivity.