All Projects

Meeting Intelligence Agent

Also on the home page

Transcribes meetings and extracts insights — decisions, action items, and open topics.

Technical Debt Analyzer

Also on the home page

Analyzes code repos and generates a 0-100 health score covering code quality, test coverage, dependencies, and CI/CD.

Document Intelligence Agent

Also on the home page

Semantic search over documents with exact citations, comparison across N documents, deployed with a production vector database.

AI Lead Capture SaaS

Also on the home page

Deployed to production on AWS EC2, with multi-tenant support and SSE streaming.

Job Matcher Agent

Also on the home page

CV-to-job-offer compatibility scored by a self-trained ML model (MAE 8.04, R² 0.629), with an AI-generated explanation report.

CrossCheck

Also on the home page

Multimodal disclosure-consistency engine: reconciles earnings-call audio, investor decks, and SEC XBRL filings for 5 real companies into evidence-linked claims. Two independent checkers — rule-based and a Claude Sonnet 5 judge — hit 97.8% accuracy and caught 2 real disclosure discrepancies.

MCP Agent

MCP server exposing 4 tools, connected directly to Claude Desktop.

AI WhatsApp Agent for Clinics

WhatsApp agent for clinics — manages appointments and patient queries via conversational AI.

RAG PDF Chat Agent

Conversational chat over PDFs with semantic retrieval, including a compare mode across two documents.

PDF Translator v2

PDF translator with a 7-node pipeline that extracts text at the span level to preserve layout and images, plus automatic quality control.

BI Agent

Business intelligence agent with automatic anomaly detection — z-score, concentration, disparity, and customer churn.

JobLens

7-step job-intelligence system with a CV advisor and a .docx CV generator.

Clipper

End-to-end pipeline for viral short-form clips (TikTok/Reels) — narrative analysis, audio-peak detection, and AI-driven visual tracking.

Taller Mecánico

Multi-agent system for running a mechanic shop — Router/Diagnosis/Quoting/Evaluator pattern with a three-way evaluation loop (approve/reject/escalate to human), an auditable decision log, PII encryption, and observability with Langfuse.

Agent Hub

Personal desktop launcher for the portfolio agents — abstraction over three connection types (external service, shared process, isolated process), lifecycle management and health checks, and its own design system.