Weekly job scraper targeting Anthropic, Netflix, Nvidia, Airbnb, and Vercel — bypasses ATS auth walls (Workday, Greenhouse, Ashby) with a multi-strategy fetch layer, scores each posting against resume via Claude API, and delivers instant Telegram DMs for high-fit matches.
System Metrics
- 5 target companies monitored: Anthropic, Netflix, Nvidia, Airbnb, Vercel
- 3 ATS bypass strategies: plain HTTP, SSR heuristic, Playwright headless
- < 60s alert latency from scrape run to Telegram DM
Key Highlights
- Multi-strategy ATS bypass: plain HTTP → SSR empty-shell heuristic → Playwright headless
- job_content bypass mode decouples scraping from scoring — works even when fetch fails
- SQLite hash-key idempotency: re-runs skip already-scored jobs, auto-detect new postings
- failure_mode taxonomy in error responses drives strategy selection and debugging
- Integrated with damilola.tech score_job MCP for consistent 0–100 resume fit scoring
Node.jsSQLitePlaywrightClaude APITelegram Bot APIGitHub Actionslaunchd
BareClaude is D's personal agentic OS — 3 autonomous principals running 24/7 on a Mac Mini. Clara Nova handles life ops, calendar, and finances. Dara Fox owns the full engineering stack (security, testing, infra, product). TARS serves as D's spouse's personal assistant. Powered by Bun fleet workers, launchd, Linear, GitHub Apps, Slack webhooks, Tailscale, and the Claude API.
System Scale
- 3 autonomous principals: Clara Nova (Chief of Staff), Dara Fox (Distinguished Engineer), TARS (Spouse's Personal Assistant)
- 20+ launchd cron jobs across planning, execution, heartbeat, PR digest, and vault sync
- Slack + Telegram + Linear + GitHub + Gmail routing via per-agent identity wrappers
Key Highlights
- 3-principal architecture: Clara (life ops + 7 domains), Dara (engineering + 9 domains), TARS (spouse's personal assistant)
- Bun fleet workers replace 19 bash daemons: launchd-supervised, keychain-accessible, no Terminal-rooted sessions
- Automated SDLC: branch → CI → CodeRabbit → 3-gate merge check → D clicks Merge
- Queue-based dispatch: cron jobs and Slack webhooks enqueue via flock-protected scripts
- Infra drift detection: daily manifest-driven check with unencrypted-secrets scanner
- Telegram inbound DMs via bun/grammy listeners — each agent has its own supervised bot
- launchd service management with KeepAlive auto-restart and health probes
- Tailscale-secured Mission Control dashboard with real-time fleet state and SQLite telemetry
TypeScriptBunNext.jsClaude APIlaunchdTailscaleGitHub AppsLinearSQLite
Complete ticketing and event management system built from scratch to handle multi-tier pricing, payment processing, and attendee check-in.
Technical Achievements
- 451K lines of code across 931 JavaScript files
- 3,229 tests at 92% coverage with 71 DB migrations
- Significant revenue processed at inaugural event
Key Highlights
- Circuit breaker pattern with automatic database failover and health monitoring
- Multi-tier caching (Redis L2 + Memory L1) with intelligent promotion
- Live scoring system with real-time WebSocket updates for competitions
- 3-tier disaster recovery: PITR (24h) + daily backups (30d) + monthly snapshots
- Apple Wallet and Google Wallet pass generation with JWT-based QR codes
JavaScriptNode.jsVercelStripeBrevoTurso
Personal portfolio with an AI chatbot that answers recruiter questions about experience, skills, and role fit using Claude with full context.
Technical Metrics
- 13K source LOC with 48% test-to-code ratio
- 34 unit tests + 6 E2E suites across 5 browsers
- 36 React components with 21 API routes (12 public + 9 admin)
Key Highlights
- Full-context LLM (no RAG) with Anthropic prompt caching for cost optimization
- Production admin portal: JWT auth, 10 pages, real-time analytics dashboard
- Distributed rate limiting: Redis + circuit breaker with graceful degradation
- Comprehensive audit system: 18 event types with IP anonymization
- Recruiter-ready resume generator: Claude-powered tailoring with PDF export
- Security hardening: CSRF, SSRF prevention, timing-safe comparisons
- UTM traffic tracking: Full attribution with source/medium/campaign analytics
- Private content submodule: Git submodule with Vercel Blob sync workflows
- Deterministic resume scoring: Reproducible algorithm with weighted criteria
Next.jsTypeScriptTailwind CSSClaude APIVercel
10 production workflows with custom CI/CD pipeline (1,634 LOC) solving Pipedream's missing deployment API. AI-powered task prioritization using Claude Opus 4.5.
Technical Metrics
- 10 workflows with 8,620 LOC across 40 Python files
- 6 APIs integrated with 60%+ test coverage enforced
- Custom deployment engine: 2,678 LOC browser automation
Key Highlights
- Built CI/CD when vendor has no API: Playwright automation with Google SSO
- AI task scoring using Claude Opus 4.5 with batch processing (40 tasks/batch)
- Production resilience: 4 different rate limit strategies across services
- Bidirectional sync: Gmail/Notion/Tasks/Calendar with idempotency guarantees
- ThreadPoolExecutor parallelization: 6 Claude workers, 10 Notion workers
PythonPlaywrightClaude APINotion APIGmail APIPipedream
Production AI orchestration framework: 12 agents (consolidated from 31) achieving 4-6x performance through multi-instance parallelization.
Framework Scale
- 107K+ lines with 80 automation scripts
- 12 agents, 20 commands, 17 skills, 85+ docs
- Substantial performance improvement via multi-instance parallelization
Key Highlights
- Strategic consolidation: 31 agents merged to 12 while maintaining full functionality
- SYSTEM BOUNDARY security architecture prevents unauthorized agent escalation
- Thinking-level calibration: ultrathink tokens (31,999) for complex reasoning
- One-command deployment (/sync) with automatic backup and rollback
- Three-tier execution model: Direct, Skills, Agents for optimal efficiency
- Composable workflows: /ship-it supports flags (-d -t -c -r -p -pr)
YAMLMarkdownPythonShellClaude CodeGitHub Actions