ASSISTANT PROFESSOR / RESEARCH SCIENTIST
Mirko Marras
Passionate expertise in responsible artificial intelligence research and development — all from the enchanting Sardinia, Italy
ABOUT MY
CAREER
I'm an inventive scientist with a Ph.D. in Computer Science and over 8 years of experience. Under my faculty position at University of Cagliari, Italy, I research and develop technology that sustains responsible AI principles — human agency and supervision, robustness and safety, privacy, transparency, diversity, equity, fairness, societal well-being, and accountability — for the benefit of individuals and society.
AMBITIOUS
VISION
Strong background in deep learning, dedicated to advancing science and technology of responsible intelligent machines for vision, language, speech, tabular data, and time series. Committed to ambitious, long-term research with ability to think scientifically, innovate, and communicate effectively both within a technical team and to non-technical peers and the general public. Deeply committed to fostering an inclusive culture that values diverse backgrounds and perspectives.
EXPERTISE
Extensive software development experience on manipulating heterogeneous data sources and large-scale computing resources to leverage data semantics for building machine learning and deep learning models. Strong expertise in data modeling, such as knowledge graphs, and methods, including supervised (classification, recognition, recommendation, retrieval), unsupervised (language modeling, generative), and reinforcement learning, with libraries like TensorFlow, applied in fields such as biometrics, education, and smart cities.
TRACK RECORD
Leadership in research and engineering, including the management and mentorship of 40+ students, focused on producing seminal and influential scientific work that advances the understanding of intelligence. Proven success through grants (1 as a Principal Investigator, 4 as a Local Unit Coordinator), fellowships (such as EPFL), and internships (such as Eurecat and NYU), as well as 90+ publications also in top-tier, peer-reviewed artificial intelligence conferences (such as AAAI, AIED, CIKM, ECIR, EDM, IJCB, INTERSPEECH, LAK, SIGIR, UMAP) and journals (such as ACM TIST; IEEE TIFS, TLT, and JSTSP; Elsevier CiHB, IF, IP&M, KBS, and PRLetters).
EXPERIENCE
& STUDIES
RECENT
NEWS
Take a look at my activities and accomplishments within responsible intelligence
EXPERTISE
& SKILLS
Blending technical know-how with people skills to shape a thoughtful vision for artificial intelligence
- FOUNDATIONAL SOFTWARE DEVELOPMENT
- DATA ENGINEERING & ANALYSIS
- ML/DL DEVELOPMENT & DEPLOYMENT
- GENERATIVE AI DESIGN & DEVELOPMENT
- RESPONSIBLE AI ANALYSIS & COUNTERING
- SPECIALIZED APPLICATIVE DESIGN & DEVELOPMENT
100 %
Proficient in several operating systems, diverse programming languages, using integrated development environments, producing documentation, and managing version control.
- -Operating Systems: MS Windows, Linux, Android
- -Programming Languages: C++, Java, Python
- -IDE: PyCharm, Visual Studio Code, DevCpp, Eclipse
- -Documentation: Sphinx, LaTeX, MS Office Suite, Google Suite
- -Version Control: SVN, GitHub, Bash
70 %
Familiar with transforming complex data into actionable insights via data science tools, big data technologies, sophisticated visualizations, and data representation methods.
- -Storing: MySQL, MongoDB, Neo4J
- -Processing: Jupyter, Pandas, NumPy, SciPy
- -Representing: Knowledge Graphs, RDF, Turtle
- -Visualization: Matplotlib, Seaborn, Bokeh, D3.js
- -Scaling: Hadoop, Spark
95 %
Expert in creating and using machine learning techniques, from shallow to advanced deep learning methods, with a strong focus on rigorous evaluation and statistical measurement.
- -Shallow ML: DT, SVM, KNN, NB, RF, GB, K-Means, DBSCAN
- -DL Methods & Architectures: CNN, LSTM, GNN
- -ML & DL Frameworks: PyTorch, TensorFlow, Scikit-Learn
- -NLP and CV Tools: Hugging Face, NLTK, Spacy, OpenCV
- -Deployment: YAML, Wandb, Gradio, Flask, Docker, ONNX
75 %
Skilled in harnessing generative architectures, advanced large language modeling, and retrieval-augmented generation to develop modern human-centered solutions.
- -Generative Networks: AE, VAE, GAN, CycleGAN, DCGAN
- -Generative Architectures: Transformer, U-NET
- -LLMs: Falcon, Gemini, GPT, LLAMA, Mistral, Palm
- -Fine Tuning: PEFT, LoRA, QLoRA, LLaMA-Adapter
- -RAG: DSI, HyDE, Multi-query, MMR, LLM rerank
85 %
Competent in applying advanced techniques for model interpretability and explainability to ensure transparency, as well as for bias analysis and mitigation to enhance fairness.
- -Fairness Notions: Independence, Separation, Sufficiency
- -Fairness Methods: Sampling, Regularization, Calibration
- -Fairness Understanding: Causality
- -Explainability Methods: Anchors, CEM, CF, IG, LIME, SHAP
- -Explainability and Fairness Tools: Alibi, AIF360, FairML
90 %
Proficient in tackling a wide range of application areas including smart city solutions, face and voice biometric analysis, educational decision support, and recommendation.
- -Education: Dropout, Warning, Knowledge Tracing, Profiling
- -Face Biometrics: Generation, Verification, Identification
- -Recommendation: Collaborative Filtering, Knowledge-aware
- -Smart City: Anomaly Detection, Object Detection and Tracking
- -Speech Processing: Recognition, Verification, Impersonation
100 %
Proficient in several operating systems, diverse programming languages, using integrated development environments, producing documentation, and managing version control.
- -Operating Systems: MS Windows, Linux, Android
- -Programming Languages: C++, Java, Python
- -IDE: PyCharm, Visual Studio Code, DevCpp, Eclipse
- -Documentation: Sphinx, LaTeX, MS Office Suite, Google Suite
- -Version Control: SVN, GitHub, Bash
70 %
Familiar with transforming complex data into actionable insights via data science tools, big data technologies, sophisticated visualizations, and data representation methods.
- -Storing: MySQL, MongoDB, Neo4J
- -Processing: Jupyter, Pandas, NumPy, SciPy
- -Representing: Knowledge Graphs, RDF, Turtle
- -Visualization: Matplotlib, Seaborn, Bokeh, D3.js
- -Scaling: Hadoop, Spark
95 %
Expert in creating and using machine learning techniques, from shallow to advanced deep learning methods, with a strong focus on rigorous evaluation and statistical measurement.
- -Shallow ML: DT, SVM, KNN, NB, RF, GB, K-Means, DBSCAN
- -DL Methods & Architectures: CNN, LSTM, GNN
- -ML & DL Frameworks: PyTorch, TensorFlow, Scikit-Learn
- -NLP and CV Tools: Hugging Face, NLTK, Spacy, OpenCV
- -Deployment: YAML, Wandb, Gradio, Flask, Docker, ONNX
75 %
Skilled in harnessing generative architectures, advanced large language modeling, and retrieval-augmented generation to develop modern human-centered solutions.
- -Generative Networks: AE, VAE, GAN, CycleGAN, DCGAN
- -Generative Architectures: Transformer, U-NET
- -LLMs: Falcon, Gemini, GPT, LLAMA, Mistral, Palm
- -Fine Tuning: PEFT, LoRA, QLoRA, LLaMA-Adapter
- -RAG: DSI, HyDE, Multi-query, MMR, LLM rerank
85 %
Competent in applying advanced techniques for model interpretability and explainability to ensure transparency, as well as for bias analysis and mitigation to enhance fairness.
- -Fairness Notions: Independence, Separation, Sufficiency
- -Fairness Methods: Sampling, Regularization, Calibration
- -Fairness Understanding: Causality
- -Explainability Methods: Anchors, CEM, CF, IG, LIME, SHAP
- -Explainability and Fairness Tools: Alibi, AIF360, FairML
90 %
Proficient in tackling a wide range of application areas including smart city solutions, face and voice biometric analysis, educational decision support, and recommendation.
- -Education: Dropout, Warning, Knowledge Tracing, Profiling
- -Face Biometrics: Generation, Verification, Identification
- -Recommendation: Collaborative Filtering, Knowledge-aware
- -Smart City: Anomaly Detection, Object Detection and Tracking
- -Speech Processing: Recognition, Verification, Impersonation