dsaavedra.dev

Daniel Saavedra

Product, Platform & Technology Leader | B2B SaaS · Industrial IoT · Cloud Operations · FinOps

Personal / Side Projects

Live demo

Elevator Maintenance Flow

AI-powered predictive maintenance for vertical transport

End-to-end PoC exploring how AI and real-time IoT data can transform elevator maintenance from reactive to predictive. A risk model surfaces the units most likely to fail, and an on-demand voice briefing — generated with Amazon Bedrock and read aloud in the browser — turns each unit's risk data into an actionable summary for technicians, helping Operations Managers cut downtime and optimise their teams. Built using Spec-Driven Development (SDD) with Claude Code — a deliberate step into the Product Maker space — with a React + Vite frontend and FastAPI + PostgreSQL backend, deployed to AWS via a GitHub Actions CI/CD pipeline (Docker, nginx, TLS).

React FastAPI Python PostgreSQL Docker AWS CI/CD IoT AI/ML GenAI Amazon Bedrock Spec-Driven Dev Claude Code
Live · Volunteer

SismoAyuda VE

Post-earthquake structural triage for Venezuela

On 24 June 2026 two earthquakes — Mw 7.2 and Mw 7.5, thirty-nine seconds apart — hit La Guaira, Caracas and Yaracuy, killing thousands and leaving tens of thousands injured or displaced. SismoAyuda VE is not my project: it was created and is led by the Hacha y Machete collective, and above all by Luken Ignacio Quintana, who stood it up in a matter of days. Citizens upload photos of a damaged building; volunteer civil engineers and architects assess them remotely against ATC-20 and EMS-98 criteria and return a habitability rating. It later grew into the tool that Colegio de Ingenieros de Venezuela (CIV) inspectors use in the field to score structural risk and record the results of official on-site inspections.

I joined as a volunteer on 28 June, four days after the earthquakes, having come across the initiative on LinkedIn. I helped set the project up in Jira and coordinate delivery through the first two weeks, and then took on the work that a build made at emergency speed necessarily defers: a full data-protection layer (GDPR + LOPDGDD with Venezuelan legal anchoring — consent capture and consent log, privacy policy, data-subject rights) and a series of security audits that translated into concrete fixes across authentication, authorisation and data integrity. The repository is private; the platform is live and in use.

Next.js TypeScript Supabase PostgreSQL Vercel Security Audit GDPR / LOPDGDD Data Protection Crisis Response Jira Volunteering