Hi, I'm Miguel
Python Developer, AI Engineer.
Experience in backend development with Python, API development, and automation. Skilled in designing and implementing AI and Machine Learning solutions.
Experience in backend development with Python, API development, and automation. Skilled in designing and implementing AI and Machine Learning solutions.
Check out some of my personal and paid projects
Description: Designed and implemented a province-level forecasting system for respiratory hospital admissions using time-series machine learning. Missing data were imputed with Random Forest models and several model families were evaluated; recurrent neural networks (RNNs) achieved the best performance ( >90% in validation metrics on the prepared datasets).
Impact: Accurate short-term forecasts that support improved resource allocation and planning in local health services.
Description: Developed an automated pipeline in Python to transform digital books into produced audiobooks and stylized graphic-novel assets. Built end-to-end tooling for TTS customization (multiple voices, speeds and tones), prompt engineering with OpenAI and Midjourney for narrative and visual style, and containerized workflows for reliable production.
Impact: Dramatically reduced manual production time and enabled repeatable, configurable chapter-level generation suitable for distribution.
Description: Implemented a hybrid OCR pipeline to convert scanned journal pages (text + mathematical formulas) into LaTeX, combining classical preprocessing (denoising, bbox refinement) with specialized ML for equation recognition (TexTeller). The system classifies regions, resolves overlapping bounding boxes, routes text to Tesseract and formulas to the deep-learning OCR, and includes human-in-the-loop verification for complex cases.
Impact: Substantial reduction in manual transcription effort and faster digital publication of mathematical content.
Description: Researched and prototyped an on-chain reputation layer to improve trust for P2P blockchain transactions. Implemented smart contracts, off-chain data aggregation and REST APIs to compute reputation metrics from on-chain behavior; tested deployments locally using Hardhat/Ganache.
Impact: Provides an extensible reputation signal that can be integrated into marketplaces or DApps to reduce counterparty risk.
Description: Built an agent-based village simulation combining SimPy for discrete-event simulation and LLM-driven decision agents to generate context-aware tasks for agents. The system models resources, stochastic events and agent behaviors; LLM prompts synthesize village state to create dynamic, scalable decisions per agent.
Impact: Enables exploration of emergent behaviors and policy scenarios where human-like decision reasoning is required.
Description: Created a production-ready, configurable CRUD REST API using FastAPI, SQLAlchemy and Alembic, with Pydantic models for strict validation and Docker for reproducible deployments. Focused on clear API design, migration flows and reusability so the codebase serves as a starter template for larger systems.
Impact: Reduces time-to-market for new services and enforces consistent validation and schema evolution.
Description: Implemented a web-scraping pipeline to collect video game data and an information-retrieval engine that blends multiple ranking strategies: TF-IDF vector search, tag-based similarity, and genre similarity. Results from the three strategies are normalized and aggregated into a unified ranking score (0–1) to produce robust search results.
Impact: Improved search relevance by combining orthogonal similarity signals and providing a defensible ranking metric.