DC / Engineering
Software engineerColombia / remote

Daniel
Castrillon

I build production software across product architecture, data, authentication, cloud delivery, and the engineering systems that keep change safe.

309k+

unique source lines audited

18

owned repositories measured

24

high / critical findings

3

production travel platforms

Systems with real operational weight.

These are not interface exercises. They combine public product surfaces with private operational systems, data, permissions, migrations, and deployment work.

01Travel platform

Mitiquete

A production travel platform spanning global discovery, booking operations, content, authentication, media, and cloud delivery.

Role
Lead software architecture and AI-assisted engineering
System
Next.js monorepo, TypeScript, PostgreSQL, Prisma, AWS S3 and SES
Scope
Product architecture, CMS, booking workflows, migrations, security, and delivery
Security
Zero-Trust RBAC and remediation of 17 critical findings across production surfaces
Production / live system
Mitiquete production website showing its global travel search experience
02Content platform

Conociendo Colombia

A tourism publishing and itinerary platform designed around structured content, arrival points, regional discovery, and reusable platform foundations.

Role
Platform architecture and full-stack engineering
System
Next.js, TypeScript, PostgreSQL, Prisma, localized content, and media workflows
Scope
Structured publishing, route discovery, administration, data design, and shared CMS components
Focus
Editorial clarity for travelers without sacrificing operational control for the business
Production / live system
Conociendo Colombia production website showing its city-of-arrival route builder

How I make complex work reviewable.

  1. 01

    Model boundaries before features.

    I start with ownership, trust, data flow, failure modes, and operational constraints. The interface comes after the system has a shape.

  2. 02

    Keep authorization close to the action.

    Authentication is not authorization. Every mutation, route, and storage path must preserve tenant, role, and ownership boundaries.

  3. 03

    Make AI output earn trust.

    Coding agents accelerate implementation, but generated changes still pass typed contracts, tests, security review, and human judgment.

  4. 04

    Ship evidence, not adjectives.

    Production behavior, reproducible measurements, reviewable changes, and explicit tradeoffs communicate more than inflated titles.

Work that can be inspected and challenged.

Codex Desktop Linux

My updater compatibility pull request adds defensive support for legacy bundle inputs and Linux package rebuilds, with regression coverage across RPM, Debian, Pacman, and Nix workflows.

Inspect pull request #991
Application security

Documented and drove remediation of 24 high-severity or critical findings across production systems, with emphasis on authorization dominance and exposed mutation surfaces.

Research communication

Delivered an invited talk at the Max Planck Institute for Security and Privacy on structural security risks in AI-assisted software systems.

Production first. Education in support.

I am a Colombian software engineer working across platform architecture, secure product delivery, and AI-assisted development. I care about systems that remain understandable after the demo is over.

Primary stack

TypeScript, Next.js, React, PostgreSQL, Prisma, Node.js

Additional work

Go, Python, C#, SQL, Linux, Docker, AWS, Vercel

Education

B.Sc. Software Engineering, expected 2027 · GPA 4.63 / 5.0