Candidate Intel Agent
Parallel research agent that compiles candidate intelligence from resumes, GitHub, publications, and social signals into leveling and comp briefs.
Projects
Flagship products first, then active systems, then supporting tools and earlier work.
Flagship
Deterministic prompt governance for production LLM systems — score, route, enforce, and audit.
Native macOS menubar widget that tracks Claude usage limits in real time. Single-file Swift app, no dependencies, Homebrew install.
An AI talent mapper that turns a job title into a full sourcing landscape in under 10 seconds.
Blind-test 100+ LLMs side by side. Real prompts, no marketing — just truth scores and failure analysis.
Personalized offer pages generated from ATS data. Turns a transactional moment into a candidate experience artifact.
Talent operating system that encodes hiring doctrine into software — scoring, JD generation, calibration, and drift detection.
Active
Interview evaluation skill that applies a 7-layer hiring doctrine to transcripts. Outputs scorecards, risk classifications, and scaling assessments.
Triage assistant for Slack, Gmail, and Calendar. Classifies work, drafts responses, tracks items to resolution.
Research and fleet management hub. Coordinates multiple agent projects, keeps docs aligned, enforces release discipline.
Fully autonomous kids video pipeline with safety gates, observability, and self-improving feedback loops.
Internal Tools
Internal tools the recruiting team uses daily.
Transcript analysis app that generates structured interviewer feedback, candidate evaluations, and PDF-ready reports against our Hiring for Great framework.
Parallel research agent that compiles candidate intelligence from resumes, GitHub, publications, and social signals into leveling and comp briefs.
Job description generator. Enforces structure, outcome focus, and consistency. Recruiters use it instead of writing JDs from scratch.
Real-time interview support interface that surfaces follow-up prompts and coverage guidance during live conversations.
Open Source
Public repos, side projects, and earlier work.
Full-stack app that scores job descriptions on clarity, inclusivity, and SEO — then suggests concrete improvements.
Early v0 prototype for structured job descriptions. Led to the later JD Genius Builder.
Natural-language hotel search agent that profiles hotel DNA and explains why each recommendation fits — not just price.
GitHub contributor extraction and enrichment tool for building candidate pipelines from open source activity.
Homebrew tap for distributing developer utilities like Claude Usage Widget with one-command install.
Team wellness check-in app. Next.js + Supabase.
Open-source arena that blind-tests 100+ AI models on real prompts from 4 Kaggle datasets. Trust scores, failure classification, BYOK support.
Private Projects
Private repos contain internal workflow, candidate data, or proprietary operating detail. Happy to walk through the architecture and decisions for any project listed here.