Ismael Garrido Muñoz

PhD - Universidad de Jaén (2021 - Current)

Bias analysis, detection and mitigation in deep learning-based language models

I am currently pursuing a PhD in deep learning Biases in Deep Learning, focusing on the development of ethical AI systems and enhancing AI transparency. My research aims to address and mitigate biases in AI models to ensure fairness and equity. Additionally, I am dedicated to improving AI interpretability, striving to move beyond the "black box" nature of deep learning algorithms. By promoting ethical standards and fostering a clear understanding of AI, my work contributes to creating more reliable and trustworthy AI technologies. To achieve these goals, I employ a combination of machine learning, natural language processing, and data visualization techniques. I develop tools and methodologies to detect and mitigate biases in deep learning models, ensuring they operate fairly and transparently.

Máster Universitario en Ingeniería Informática - Universidad de Jaén (2016 - 2018)

TFM: Confirmation bias analysis from massive social media data.

Grado en Ingeniería Informática - Universidad de Jaén (2012 - 2016)

TFG: Follow-up of press topics

Natural Language Processing Specialization (Coursera)

Classification and Vector Spaces, Probabilistic Models, Sequence Models, Attention Models.

Oracle OCI AI (Oracle)

Oracle AI Vector Search Certified Professional, Oracle Cloud Infrastructure 2025 Certified AI Foundations Associate, Oracle Cloud Infrastructure 2025 Certified Data Science Professional, Oracle Cloud Infrastructure 2025 Certified Generative AI Professional.

Senior Developer en Liquid Barcodes Liquid Barcodes (2022 - 05/2025)

Full stack development. API as a Service/SASS for an advertisement/loyalty platform for the retail sector, using microservices (C#, RabbitMQ, Mongo, Redis, AWS) along with some other produts and client side apps (Angular/Vue/React) that either serve as a frontend or are used to manage the advertisment campaigns. Worked on CICD pipelines (Jenkins/Github Actions) within a Kubernetes Cluster, data (mostly Tableu/python), internal tooling.

Full Stack developer at Diputación de JaénUD Ibérica (Abril 2020 - 2021)

Responsible for multiple web services and apps: Sede Electrónica, Acceso (Oauth like handling extra login methods & integrations available to Spanish Citizens), VideoActas, Contro Horario, Plan Director Web (back-office) with .NET related tecnologies, Vue, Angular, Oracle/Informix/SqlServer. Developer of a custom frontend framework integrating the legacy MVC with the new MVVM to accelerate the development of new apps.

Cloud Developer at OfimaticaOfimatica TSS (Julio 2019 - Abril 2020)

Developed a comprehensive cloud-based SaaS solution tailored for the hotel industry covering the full business vertical of hotel operations, utilizing a multi-tenant architecture and CQRS (Command Query Responsibility Segregation) pattern. Worked on performance improving scalability of the system. (C#, Angular, Azure, Azure Queues, MongoDB)

Researcher at the sinbad2 group sinbad2.ujaen.es (Enero 2018 - Julio 2019)

Development of Flintstones 4 modular software (Fuzzy LINguisTic DeciSion TOols eNhacemEnt Suite) for running multi criteria decision making algorithms. Used by research groups. Implemented multiple research papers into it. (Eclipse RCP/Java 8, Web SVG for custom charts)

From Body to Mind: Analyzing Gender Representation in Spanish Generative Language Models. (to be published)

Analyzes gender bias in 20 Spanish LLMs, finding that models describe women with physical and emotional attributes, while associating men with behavioral and cognitive traits, reflecting cultural stereotypes.

Towards Reliable AI Fairness: Challenges in Steering Features within Bias-Implicated Neurons. (to be published)

Investigates 'feature steering' for bias mitigation. The study finds the technique is highly sensitive to parameters and context, making it challenging to apply as a reliable, large-scale solution.

Agentic Personas: An Interactive Tool for Natural Language Configuration of Persona Generation. SEPLN 2025

Presents 'AgenticPersonas,' an interactive tool using a conversational agent to help users configure persona generation. This facilitates the 'tag-first' approach to mitigate bias in AI evaluation.

Tag-First: Mitigating Distributional Bias in Synthetic User Profiles through Controlled Attribute Generation.

Introduces a 'tag-first' framework to generate diverse personas for AI bias testing. This method separates attribute selection from narrative generation to mitigate LLM bias and enable controlled fairness evaluations.

Precise bias location in Language Models

This work addresses how Large Language Models (LLMs) like GPT-4 and Gemini perpetuate biases due to their statistical, non-comprehension-based nature. It proposes a new technique to precisely detect where these biases are located within the model's hidden states. The method is initially developed on smaller models, with the aim of scaling it to larger ones.

Exploring gender bias in Spanish deep learning models (Read)

This work presents a visualization tool designed to investigate gender bias in Spanish language models. The tool uses template sentences with male or female contexts to compare model outputs. It allows users to explore these biases at different levels of detail, from specific model weights to aggregated results.

Analysis, Detection and Mitigation of Biases in Deep Learning Language Models (Read)

This paper introduces the problem of societal biases (based on gender, race, religion, etc.) being learned and amplified by deep learning models. It frames this issue within the research fields of AI Fairness and Explainability. The work analyzes how these biases manifest and discusses methods for their detection and mitigation.

A Survey on Bias in Deep NLP (Read)

This paper is a survey that reviews the issue of bias in deep learning models for Natural Language Processing (NLP). It provides a formal definition of bias in this context. It then explores and summarizes the different methods researchers have developed to detect and correct biases in NLP networks.

Deep Learning, Data Science, NLP

• Natural Language Processing & Deep Learning: Proficient in SpaCy, Scikit-learn, NLTK, Hugging Face, and PyTorch. Experienced with both mask-based (e.g., BERT) and generative models (e.g., GPT-2). Familiar with Large Language Models (LLMs), such as LLaMA and Mixtral, and utilizing LLMs as a service through platforms like Groq and OpenAI.
• Topic Modeling & Classification: Expertise in MALLET, LDA, and BERT-based models, including RoBERTa, for advanced topic modeling and text classification.
• Data Visualization & Analysis: Skilled in data visualization using Matplotlib, Chart.js, Graphviz, Seaborn, Kibana, and Grafana. Experience in data analysis and Business Intelligence (BI) with custom tooling, Tableau, and KNIME.
• Tooling: Proficient in Slurm and Vast.ai for job scheduling and distributed computing.
• Web Crawling & Scraping: Extensive experience with web crawling and scraping using PhantomJs, Playwright, and custom tooling.
• Other: Competent in Text-to-Speech (TTS) using Whisper and custom data preprocessing tools.

Senior Full Stack

• Frontend Development: Proficient in JavaScript/TypeScript (Vue, Angular, jQuery, Vanilla JS) and CSS/SaSS (Bulma, Bootstrap, Tailwind). Skilled in HTML markup with microdata for enhanced accessibility, search engine optimization, and semantic web. Experience with WebAssembly.
• Backend Development: Proficiency in C#, PHP, and Python frameworks, including .NET MVC, ASP.NET Core, Flask, Laravel, and FastAPI.
• Cloud Services: Experience with AWS services (EC2, Aurora, DocumentDB, Elastic Cache, SNS, SQS, S3) and message queuing systems like RabbitMQ and Azure Queues.
• DevOps & Tooling: Expertise in bundlers such as Vite, Webpack, and Laravel Mix. Extensive experience with CICD pipelines, primarily using Jenkins, GitHub Actions, and ArgoCD. Skilled in containerization technologies, including Docker and Kubernetes.
• Database Management: Strong knowledge of SQL databases (MySQL, MariaDB, Oracle, Informix, SQL Server, SQLite) and NoSQL databases (MongoDB, Redis).

Infinite Craft browser only clone

Clone of Infinite Craft running entirely in the browser using a RAG like architecture running on WebAssembly. On entering the site will download a sentence-transformer model from HuggingFace and create a HNSW index, this way the Language Model will fully run on your device. All generation and inference is done client side. Takes some minutes to load. https://craft.ismael.codes

RAE explorer

RAE explorer is a tool to explore the Real Academia Española (RAE) dictionary fully in your browser using sql.js compiled to WebAssembly and a tiny sqlite database. https://rae.ismael.codes

Background Remover

Background remover tool that runs fully in your browser, privacy garanteed. Uses a tiny model and transformers.js for fast inference (as fast as the browser can go). https://bg.ismael.codes

WASM OCR

WebAssembly based ORC tool. https://ocr.ismael.codes

WASM Hash

Hash calculater with WASM. https://hash.ismael.codes

WASM Watermark remover using LaMa

Remove watermarks from images direcly in the browser. https://watermark.ismael.codes

WASM Text to Speech

Using the Supertonic ONNX model. https://tts.ismael.codes

Fractured JSON formatter

JSON formatter for humans. https://json.ismael.codes