Jose Nobile — AI Engineering Profile
I build payment systems, decompose monoliths, harden security, and deploy production AI systems. Full-stack platform engineer owning outcomes end-to-end across 29 microservices, with the OpenClaw AI gateway (Claude Opus 4.6 + GPT-5.4 failover) running 24/7 on Docker.
Solutions I’ve Built
Real problems solved in production. Each card: what was broken, what I built, and the measurable impact.
Multi-Country Payment Infrastructure
Problem: No unified payment flow, manual onboarding, missing fraud prevention, no tax compliance across Latin America.
- Built complete Stripe Connect Express integration: onboarding, fee management, connected accounts
- Stripe Checkout migration with 3DS enforcement
- No-show fee system spanning 10+ repos (Check-In, Android, iOS, Web, Payments, Lessons, Reports)
- Tax compliance: Chile (SII/Boleta Exenta), Peru, Colombia (Siigo electronic invoicing)
- Fraud prevention: stolen card detection + payment blocking
- Production incident response: payout rescue, 3DS failures affecting multiple studios
Impact: End-to-end payment infrastructure serving multiple countries with automated tax compliance and fraud prevention.
Platform Security Hardening
Problem: No rate limiting, SQL injection vectors, hardcoded credentials, no session-based auth.
- Redis-backed sliding-window rate limiting across 29 microservices (Lua scripts, cluster-safe)
- Cloudflare edge rate-limit blocker with expiry sweeper
- SQL injection remediation (parameterized queries)
- Removed hardcoded credentials from codebases
- Built session-based dynamic authentication with Authorizer service (from scratch)
- Geolocation-based login authorization
Impact: Zero successful attacks after hardening. Defense-in-depth across edge, application, and database layers.
Monolith → 26 Microservices
Problem: Monolithic backend with tight coupling, single point of failure, unmanageable deployments.
- Decomposed a monolithic backend into 26 independently deployable services
- Each service: Dockerfile, Helm chart, GitLab CI pipeline, health endpoints, K8s resource tuning
- Queue architecture with ZeroMQ for async processing
- MySQL connection resilience across all services (auto-reconnect)
- Fixed connection pool exhaustion causing OOM crashes in production
Impact: Independent deploys, isolated failures, horizontal scaling per service. Zero-downtime releases.
Webhook Delivery Engine
Problem: No way for third-party integrations to receive real-time platform events.
- Built from scratch: idempotent delivery, exponential retries, HMAC signing, multi-config fan-out
- Management API + Admin UI with event log and developer docs
- 10+ event types across the entire platform
Impact: Enabled third-party integrations with reliable, secure, auditable event delivery.
CI/CD & Infrastructure
Problem: Manual deployments, no container orchestration, slow builds, OOM failures in production.
- Kubernetes on GKE with Istio service mesh
- Docker to Kaniko migration across 10+ repos
- Helm charts for every microservice
- White-label app automation: 41 branded Android apps from JSON config
- OOM prevention (memory tuning), GeoLite2 caching (fixed 504s)
- GitLab Runner (concurrent: 10)
Impact: Fully automated pipeline from commit to production. 41 branded apps built from a single config.
Full-Stack Delivery
Problem: Slow admin panel, fragmented frontend stack, no email analytics, missing mobile features.
- Angular admin panel: lazy loading overhaul, 730 KB savings, Lighthouse 90+
- React + Vue.js consumer apps with Stripe Checkout integration
- Android + iOS native: no-show fees, coupon support, biometric check-in
- Email infrastructure: send/delivery/bounce analytics, 10 min to 10 sec reporting
Impact: Admin panel 730 KB lighter. Email reporting 60x faster. Native apps with complete feature parity.
Accomplishment Timeline
30 high-impact accomplishments across 3+ years. Each card represents a meaningful production milestone.
30 milestones • Dec 2022 – Apr 2026 • Auto-scrolling
AI-Augmented Engineering
AI is not a buzzword in my workflow. It is a force multiplier I use daily to ship faster and with higher quality.
8 Concurrent Claude Code Instances
Ran overnight for tax billing verification across 3 countries, 7 repos, 1,740 tests, 11 commits. One engineer, eight parallel AI agents.
8 Custom Claude Code Skills
/testing, /deploying, /debugging, /developing, /mr-review, /jira-docs, /planning, /cli-tools. Each skill encapsulates domain-specific workflows for the codebase.
2,205 Claude Code Sessions
Across 26 active projects. Not just chat prompts: structured agent sessions for code generation, review, testing, and deployment. 480 commits in the last 30 days alone.
OpenClaw AI Gateway
Docker container with Claude Opus 4.6 + GPT-5.4 failover. Multi-channel AI assistant (WhatsApp + Telegram) with voice transcription, vector memory (sqlite-vec), and billing proxy routing through subscription.
OpenClaw Deployment Portfolio
A real production AI system I built and operate 24/7. Not a demo. Not a prototype. Running on my hardware right now.
AI Gateway & Model Failover
- Docker container with Claude Opus 4.6 primary + GPT-5.4 automatic failover
- Billing proxy routes through subscription instead of API credits
- Model resilience: automatic fallback on rate limits, errors, or outages
- Vector memory with local embeddings (sqlite-vec) for persistent context
Multi-Channel Messaging
- WhatsApp + Telegram channels with 24/7 availability
- Local Whisper STT for voice transcription (CUDA-accelerated)
- LuxTTS cloned voice synthesis (not cloud APIs)
- Self-hosted Telegram Bot API for large file handling
Remote Browser Control
- Chrome DevTools Protocol (CDP) from WSL2/Docker
- AI-driven web navigation, form filling, screenshot capture
- Puppeteer automation for banking, scraping, monitoring
IoT & Health Automation
- Xiaomi Scale BLE health tracking via systemd daemon
- Auto-deploy dashboard with body composition data
- 375+ measurements, 30+ metrics, 14+ months continuous
Cron Jobs & Services
- 9 automated cron jobs (watchdog, health deploy, currency rates, backups, etc.)
- 6 custom systemd services running 24/7
- Bank monitoring: headless Puppeteer scraping every 15 min
- Self-healing watchdogs with automatic restart on failure
Infrastructure Stack
- Docker Compose orchestration (rootless)
- 10,159 source files across the OpenClaw workspace
- GitLab Runner for CI/CD (user-mode)
- RTX 5090 GPU for local AI inference (Whisper, TTS, Vision)
Autonomous Systems I Built
Not scripts. Autonomous systems with watchdogs, failure recovery, and continuous operation.
Xiaomi Scale BLE Body Composition Tracker
- BLE health tracking via xiaomi-scale-daemon systemd service
- ADB + uiautomator scrapes Xiaomi Home app automatically
- 375 measurements over 14 months, 30+ body metrics tracked
- Auto-syncs to Google Sheets + interactive web dashboard
- Watchdog + health alerts with Windows sound notifications
Why it matters: IoT integration, reverse engineering, daemon reliability.
Banking Monitor
- Puppeteer Extra with Stealth Plugin bypasses Imperva WAF
- Local vision AI (Qwen2.5-VL-7B) solves hCaptcha autonomously
- Headless Puppeteer scraping every 15 min from Colombian corporate banking
- Cookie persistence, failure recovery, watchdog
Why it matters: browser automation at production scale with AI-powered CAPTCHA solving.
OpenClaw AI Gateway
- Docker container: Claude Opus 4.6 primary + GPT-5.4 automatic failover
- WhatsApp + Telegram with local Whisper STT + cloned voice TTS
- Remote browser control via Chrome CDP from WSL2/Docker
- Vector memory with local embeddings (sqlite-vec), billing proxy routing
- 9 cron jobs, 6 systemd services, self-healing watchdogs
Why it matters: full production AI gateway with model failover, local inference, and autonomous operation.
Technical Depth
Hardware & AI Investment
MSI Raider 18 HX (2025)
- RTX 5090 — 24 GB GDDR7
- 64 GB DDR5-6400
- 6 TB NVMe (PCIe 5.0 + PCIe 4.0)
AI Subscriptions
- $257/month (Claude Max, ChatGPT, Gemini, Perplexity)
- 6 local AI models on RTX 5090
- 6 custom systemd services always running
Total Monthly AI Investment
Published Knowledge
Consulting & Training
Remote sessions via video call. From Claude Code fundamentals to production AI agent architecture.
Payment: Binance (USDT) • PayPal • Credit/Debit card link • Prepaid packs