Pablo – React, Typescript, Node.js

A Japanese reading app designed for intermediate learners. Instead of flashcards or native content that may be too advanced, the system generated personalized passages tailored to the learner’s vocabulary level, weak areas, and knowledge gaps. Words being learned reappeared multiple times in varied contexts to reinforce retention. There were no decks to manage or backlogs to clear. Users could start reading immediately, while the system continuously surfaced relevant vocabulary based on their learning needs.

• Developed the complete product across business, concept, and technical areas.
• Led concept ideation, including business model and GTM strategy.
• Designed UI/UX, including progressive hinting and guilt-free progression tracking.
• Generated a word database by mixing and augmenting open-source options.
• Developed a content-generation algorithm to maximize learning.
• Built a prompt-generation workbench to experiment with prompts and optimize token usage.
• Implemented a performant Node.js backend, including calling a Rust integration via Node APIs for passage tokenization.
• Managed cost-effective deployment in Railway to support development needs.

The product aims to bring intelligence to software systems under uncertainty. Pablo was part of the team that developed the "Model Editor," which allows researchers to create models using a graphical UI that translates UI components into the complex proprietary format for use in the backend.

Integrating a new model type into the UI presented significant challenges: designing validation for graphical components, creating intuitive interactions, preventing invalid structures, and understanding research requirements without being a researcher.

Nevertheless, Pablo took on the bulk of this integration and delivered it successfully:

• Led the end-to-end integration of a new model type into the UI.
• Designed UX patterns for complex multi-model interactions (highlighting, snapping, validation).
• Collaborated directly with researchers to gather requirements and validate design decisions.
• Made rapid design choices in the absence of a dedicated designer, validating through team demos.
• Architected a plugin system that made future model integrations significantly faster and easier.

Data Visualization Platform for Sales.

Indeed needed to provide its Sales Reps with faster, smarter access to data. It included migrating away from legacy, clunky systems to a streamlined UI with standard data querying.

The project was split into three main pieces:

• The Data Query service: a deep back-end pulling from AWS Athena.
• The App Backend: housing all the business entities, logic, and query proxying methods.
• The App Frontend: providing visualization-rich UIs for analysis and pattern/intelligence finding.

Pablo's responsibility was maintaining and expanding the backend's feature set. He was able to eliminate data querying inconsistencies (Critical for sales to have accurate data!), reduce service latency and startup times, and make the app more resilient by helping it get to a healthy coverage, which enabled more refactors and improvements.

• Stabilized data accuracy by syncing schemas to the provider's OpenAPI spec with automated type-safe tooling.
• Eliminated latency spikes in advertiser retrieval: implemented concurrent SQLite access via Node child processes.
• accelerated deploys and rollbacks by reducing server startup 90% (300s → 30s) via multipart S3 downloads using Node.js streams.
• Enabled safe backend refactors by raising test coverage to 82% through co-implementing the DI pattern.

Real-time 3D visualization of Inventory hardware status.

An R&D project that meshed hardware and software to create a smart solution for Inventory, trying to use space in the most efficient way.

Pablo was involved in:

• Visualization Frontend: the 3D view using ThreeJS. Panning, zooming, and clicking enabled inspection at granular levels across thousands of hardware components.
• Backend: a Backend-for-frontend Nest.js app. Could query for existing Racks and provide necessary information to the Frontend. Would handle subscription and piping status and updates from the hardware through a message queue.
• Message Queue: to handle updates from the hardware, they subscribed to a RabbitMQ instance and acted accordingly.

The project was cut due to COVID-related budget cuts, but Pablo gained valuable experience in 3D work and worked alongside mechanical and electrical engineers.

• Enabled real-time monitoring at 10,000+ object scale via coordinate tracking, virtualization, batched API calls, and mesh instancing for re-use.
• Gave team confidence in 3D view by re-architecting for testability: separated visualization logic so tests could run as standard React components.
• Improved team agility and debugging by designing a DI-based architecture; collaborated across microservices and message queues.

Property Editor.

For rental properties to be rented, Owners need to tell people what amenities their property offers. Pablo was part of the team working on the property editing side of the supply.

Pablo completed several initiatives here:

• Modernized a legacy Java system to a React / React Native / Node / GraphQL stack, delivering the project three months ahead of the target date.
• Increased property information completion rates through multiple initiatives, including “Smart Cards” and the addition of new amenities as they became available.
• Following the merger with Expedia and its systems, contributed to several platform unification efforts, including co-architecting an Apollo Gateway shard (at a time when stitching was used instead of federation) and building reusable React components for multiple frontends.

• Helped drive a 10% increase in property amenity adoption by leading development on the feature, promoted to Senior (2020).
• Unified data access for property management by co-architecting the Apollo Gateway shard.
• Built reusable React components for multiple frontends.
• Accelerated legacy sunset by 3 months by building configuration-driven React/Node patterns, among other things.