Software Engineer & AI Researcher

About Me

I am a Computer and Software Engineer working as a Software Engineer at Zapia AI, and as an Artificial Intelligence lecturer and researcher at Universidad ORT Uruguay.

My work focuses on building AI-driven solutions: integrating large language models into real products, Retrieval-Augmented Generation, and serverless architectures. On the research side, I work on the verification and explainability of intelligent systems with learning components. My thesis, “Analysis, Evaluation and Improvement of Active Regular Inference Algorithms for Neural Sequence Acceptors”, received the Best Thesis Award (Systems Category) from the National Academy of Engineering.

I care about bridging research and production, and about mentoring the engineers and students I work with.

📫 alejogaratd@gmail.com

LinkedIn

GitHub

Work Experience

Software Engineer @ Zapia AI (Nov 2025 - Present)

• Build AI-powered features and backend services for a consumer AI assistant, integrating large language models into user-facing workflows at scale.

AI Researcher @ Universidad ORT Uruguay (Jun 2024 - Present)

• Research verification methods for intelligent systems with learning components.
• Develop tools to verify Large Language Models in industrial applications.
• Build chatbot applications using Retrieval-Augmented Generation (RAG) and GraphRAG techniques.

Lecturer @ Universidad ORT Uruguay (Aug 2021 - Present)

• Lead weekly Artificial Intelligence practical courses utilizing Python to teach practical lessons on subjects like Q-learning and Minimax.
• Instruct students in Data Structures and Algorithms courses, covering both theoretical analysis and C++ programming.
• Deliver courses in Haskell for Logic and Discrete Mathematics.
• Member of the Artificial Intelligence chair, mentoring students and developing course content.

Software Engineer @ Sagitta (Nov 2024 - Nov 2025)

• Co-developed AI-driven solutions across multiple projects, integrating large language models, machine learning techniques, and data processing frameworks to optimize business workflows.
• Designed and implemented web scraping pipelines to gather and preprocess data for model training and analysis.
• Built and maintained CI/CD pipelines with GitHub Workflows to automate testing, deployment, and monitoring.
• Mentored and onboarded new engineers, fostering knowledge sharing and best practices.

Founding Engineer @ Stealth Startup (Mar 2024 - Oct 2024)

• Built a platform integrating popular construction tools to streamline and consolidate data collection.
• Responsible for the development of both the backend and frontend of the web application.
• Designed and implemented solutions using technologies such as Golang, PocketBase, Flutter, and Svelte.

Backend Developer @ SHOWX (May 2023 - Feb 2024)

• Led the gradual migration of data from a NoSQL database to a SQL database.
• Responsible for designing the database tables, developing functionalities, and deploying them.
• Technologies used include Golang, JavaScript, PostgreSQL, PostGIS, Firebase, and Google Cloud.

Skills

• AI & Machine Learning: Large language models (LLMs), Retrieval-Augmented Generation (RAG), GraphRAG, neural networks, exploratory data analysis (EDA).
• Cloud & Serverless: Google Cloud (Cloud Run, Pub/Sub, Cloud Scheduler), AWS (Lambda, S3), event-driven serverless systems.
• Languages: Golang, Python, JavaScript, C++, Haskell.
• Databases: PostgreSQL, PostGIS, Cloud Firestore, golang-migrate, Alembic, data modeling, geospatial data.
• CI/CD & Data: GitHub Workflows, Cloud Run Jobs and Workflows, Scrapy, Beautiful Soup, custom extraction pipelines.

Education

• B.S., Computer Engineering

  Universidad ORT Uruguay (May 2024), final grade 93/100

• B.S., Software Engineering

  Universidad ORT Uruguay (Mar 2023), final grade 93/100

Honors & Awards

• Best Thesis Award (Systems Category), National Academy of Engineering (2024)
• Transformer + RNN: Taysir Challenge, Taysir (2023)
• C2 Proficiency (Level C1), Cambridge Assessment English (2018)

Projects

GraphRNNAutomaton

Utilized a GraphRNN architecture to generate Deterministic Finite Automata (DFA) with specific properties. Built with Python and PyTorch.

View on GitHub

RAG Chatbot

The RAG Chatbot is an application that utilizes the RAG (Retrieval-Augmented Generation) architecture to provide responses to user queries. The RAG architecture enhances the capabilities of large language models (LLMs) by augmenting their knowledge with additional contextual data. This chatbot is implemented using FastAPI, a modern web framework for building APIs with Python. It leverages the LangChain library, which provides tools and utilities for natural language processing tasks. Specifically, it uses LangChain’s Runnable interfaces to orchestrate the RAG architecture.

Neural Checker

Its main goal is to provide implementations for the structures needed for working in the Model Extraction Framework and enable the explainability and checking of complex systems in a black box context. It is developed by the Artificial Intelligence and Big Data team of Universidad ORT Uruguay.

Publications

1. Matías Carrasco, Franz Mayr, Sergio Yovine, Johny Kidd, Martín Iturbide, Juan Pedro da Silva, and Alejo Garat. “Analyzing constrained LLM through PDFA-learning,” arXiv preprint arXiv:2406.08269, 2024. Formal Languages and Automata Theory (cs.FL). https://arxiv.org/abs/2406.08269.
2. Franz Mayr, Sergio Yovine, Matías Carrasco, Alejo Garat, Martín Iturbide, Juan da Silva, and Federico Vilensky. “Results of Neural-Checker Toolbox in Taysir 2023 Competition,” in Proceedings of 16th edition of the International Conference on Grammatical Inference, eds. François Coste, Faissal Ouardi, and Guillaume Rabusseau, vol. 217, pp. 295-298, PMLR, Jul 10-13, 2023. https://proceedings.mlr.press/v217/mayr23b.html.