Assistants over internal systems
Language models wired to internal APIs through MCP: OAuth 2.0 with PKCE, role-based access control, and personal-data filtering before anything reaches the model.
I design AI infrastructures where privacy is not optional: OpenClaw orchestrates, Ollama runs locally, and MariaDB stores your data with zero exposure.
> stack status
[OK] Gateway Active
[DB] Data Layer Connected
[AI] Ollama Instance: Ready
_
Systems in production, not prototypes. Each solves a concrete problem and has been running for months.
Language models wired to internal APIs through MCP: OAuth 2.0 with PKCE, role-based access control, and personal-data filtering before anything reaches the model.
KYC flows with their own admin panel, role management and access auditing, aligned with the data-protection laws of Panama, Chile, Peru and Colombia.
Quantitative strategies on Interactive Brokers and Alpaca, with cointegration analysis, automated execution and real-time alerts.
Language models, speech-to-text and text-to-speech running on my own servers. No data sent to third-party clouds, no per-token billing.
The tools behind everything above.
Systems currently running in production. Each one requires an account; these links open their sign-in page.
Trademark docket monitoring: files, legal representatives and automatic alerts before an expiry date.
Personal budgeting on the 50/30/20 method: accounts, recurring records and period tracking.
Self-hosted chat interface over locally served language models, with speech-to-text and voice output.
Self-hosted workflow automation: scheduled jobs and integrations wiring the internal services together.
Multi-channel AI gateway: language models reachable from messaging channels, with search and persistent memory.