Systems engineer, researcher, professor and trainer — Mexico.

I study complex and opaque systems, reconstruct how they work and turn that knowledge into independent, portable platforms.

ReactVision builds infrastructure for visual and spatial computing.

Our work includes:

• Maintaining and extending the ViroReact ecosystem
• More than 130,000 annual package downloads
• Native Android, iOS and React Native integrations
• Computer-vision infrastructure
• SLAM and spatial tracking
• A cloud-anchor stack for persistent, cross-device AR sessions without depending on Google Cloud Anchors
• Work around ViroCore, RVCA and native rendering infrastructure

The objective is not only to build AR applications, but to develop the underlying perception, persistence and rendering systems they depend on.

A self-hosted systems programming language combining native performance, explicit memory control and scripting-language immediacy.

Eskiu compiles to native code through LLVM, but can also run source files directly through eskiuc run. The same language can be used for command-line applications, network services and freestanding bare-metal software.

Current capabilities include:

• A fully self-hosted compiler: lexer, preprocessor, parser, type checker and code generator are written in Eskiu
• Three-stage bootstrap fixpoint verification
• Bounded generics and structural interfaces
• Sum types and exhaustive pattern matching
• Closures and first-class functions
• async / await and event-driven concurrency
• Explicit allocation with defer and errdefer
• Slices, checked nullable pointers and opt-in runtime safety
• Direct C interoperability, including external functions and global variables
• Native HTTP/1.1 and HTTP/2 support with TLS
• A growing standard library including collections, networking, regular expressions, sorting, URLs, UUIDs, time and random-number generation
• Native and cross-compilation through LLVM

Supported targets include:

• macOS and Linux
• Windows x86-64
• ARM64 bare metal
• Nintendo 3DS through the ARMv6 hard-float backend

An ARM64 kernel written in Eskiu boots in QEMU without libc, while the hosted runtime can serve HTTP/2 over TLS.

An independent document-intelligence and biometric runtime reconstructed through reverse engineering and clean-room reimplementation.

The project began as an effort to understand how commercial identity-analysis engines represent documents, execute OCR pipelines, combine machine-readable evidence and make biometric decisions. That knowledge was then rebuilt as a portable runtime with explicit and inspectable behaviour.

Capabilities include:

• Document-template classification
• Perspective rectification
• Declarative document layouts
• Visual OCR with locale-specific routing
• ICAO MRZ parsing and validation
• Barcode and machine-readable data extraction
• Front/back and multi-page document fusion
• Document capture-risk signals
• Face detection and landmarks
• Head-pose and image-quality analysis
• Passive facial liveness
• Face embeddings and comparison
• Portable biometric templates
• Document portrait-to-selfie verification
• ROC, AUC, EER and threshold-evaluation tooling
• REST API, Python API and command-line interfaces
• ONNX Runtime inference and Docker deployment

The runtime deliberately distinguishes structural consistency, capture risk, facial similarity and liveness from legal document authenticity.

Reverse engineering and clean-room reimplementation of the cryptographic pipeline embedded in the QR codes of Mexican voter credentials.

The original implementation was contained in a closed ARM64 native library and combined seven processing layers, including:

• AES-256-CBC
• Chained 8,192-bit RSA operations
• NEON/SIMD key derivation
• ECDSA-related processing
• Proprietary binary packing and unpacking

I reconstructed the pipeline by loading and instrumenting the original binary with Unicorn Engine, intercepting native cryptographic operations and capturing intermediate state.

The resulting implementation no longer depends on the original native library and has been ported to:

• C
• Python
• WebAssembly
• Android/Kotlin
• iOS/Swift
• Go
• Eskiu

The project includes native libraries, mobile SDKs, browser execution and HTTP services, all validated against the captured reference output.

Security analysis of Mexico's national digital-signature infrastructure, used by approximately 27 million people.

The research covers:

• Deprecated cryptographic primitives
• Short salts
• Weak password-based key derivation
• Private-key protection
• GPU-assisted attack feasibility
• Proof-of-concept recovery pipelines
• Reproducible performance benchmarks

Co-authored with F. Santander Baños and J. H. Sossa Azuela at UAEH and IPN.

Research and production work on deep-learning-based identity validation, achieving approximately 99% accuracy in its evaluated deployment context.

University professor teaching:

• Artificial Intelligence
• Data Structures
• Algorithms
• Systems Programming

I focus on helping students understand how systems behave internally rather than only teaching them how to consume existing abstractions.

Personal trainer.

My current areas of study include:

• Strength training
• Weightlifting
• Biomechanics
• Sports nutrition
• Training-volume and frequency management
• Progressive overload

The methodology is similar to engineering: observe, measure, form a model, test it and iterate.

Compilers Reverse Engineering Native Systems Computer Vision Spatial Computing LLVM ARM64 WebAssembly Biometrics AR/VR Security Research

Premio a la Innovación Gubernamental
Gobierno del Estado de Hidalgo