Table of Contents

Work with me: Probabilistic AI, funding routes, and mathematical rabbit holes #

I am an Assistant Professor at the University of Coimbra working on Probabilistic AI: Bayesian modelling, prior predictive analysis, GFlowNets, amortized inference, uncertainty-aware machine learning, trustworthy adaptive systems, and the occasional mathematical rabbit hole that looked harmless at first.

I currently do not have a large funded project with many open salaried positions. What I can offer is different: a strong research environment, active international collaborations, experience with proposals, and a willingness to help excellent candidates build their own funding route through FCT, MSCA, CAPES, or related schemes.

I am fairly laid back in style, but I take the mathematics seriously. Expect friendly meetings, too many diagrams, and occasional conversations that begin with “this should be simple” and end with a new theorem-shaped problem.

Good fit: self-driven candidates who like mathematical formulation, experiments, writing, and probabilistic thinking.

Not ideal: generic “I want to do AI” emails with no research fit, no funding route, and no concrete question.

Short version #

You may be a good fit if you are interested in:

probabilistic AI, Bayesian modelling, uncertainty, or machine learning theory;
GFlowNets, amortized inference, probabilistic programming, or structured generative models;
prior predictive analysis, Bayesian workflow, tensor/matrix factorization, or latent-variable models;
trustworthy adaptive AI, machine unlearning, constrained learning, or evaluation of foundation models;
AI for entertainment, creativity, simulation, games, recommender systems, or co-creative systems.

You are especially welcome if you are self-driven, like mathematical formulation, can write clearly, and are willing to iterate between theory, experiments, and paper writing.

You are probably not a good fit if you want a fully pre-packaged thesis, only need a supervisor signature, or want to “do AI” without a more specific research question.

My research style #

My work is usually somewhere between mathematics, probabilistic modelling, and practical machine learning. I like projects where we can:

define a problem clearly;
write down the mathematical structure;
identify what is actually new;
build a prototype or simulation;
write the result as a serious research contribution.

I enjoy working with students who are curious, honest about what they do not understand, and willing to write things down. Research is often confusing before it is beautiful. The trick is to keep enough structure that the confusion becomes useful.

Good meetings often involve a diagram, a counterexample, and the sentence: “wait, this is actually a paper.”

I am relaxed about style, less relaxed about unclear notation.

I like students who enjoy both building things and asking why the thing they built has no right to work.

Research areas #

Probabilistic AI workflow #

Bayesian workflow, prior predictive checks, posterior predictive criticism, simulation-based calibration, probabilistic programming, LLM-assisted modelling, and modern workflows for building probabilistic AI systems.

GFlowNets and amortized inference #

Generative Flow Networks, amortized samplers, sequential Bayesian inference, structured distributions, causal discovery, probabilistic programming, and learning distributions over complex objects.

Priors, tensors, and latent-variable models #

Prior predictive rank selection, Bayesian tensor and matrix factorization, algebraic structure, identifiability, probabilistic circuits, and latent-factor models.

Trustworthy and adaptive AI #

Machine unlearning, continual adaptation, uncertainty, constrained learning, evaluation of foundation models, multilingual reasoning, and probabilistic approaches to robustness.

Probabilistic AI for creative and interactive systems #

AI for entertainment, co-creativity, recommender systems, simulation, games, narrative systems, agent-based modelling, and social simulation.

General advice for prospective students and postdocs #

Before contacting me, try to answer:

What kind of research question do you want to work on?
Which part of my work connects to that question?
What funding route are you considering?
What evidence do you have that you can work independently?
What do you want to learn from working with me?

A good first email does not need to be perfect. It should be concrete.

A bad first email is usually generic: “Dear Professor, I am interested in AI, please find attached my CV.” I receive variants of this often. The sacred autocomplete spirits are powerful, but they are not a research plan.

Skills I value #

Independence #

Can you start shaping your own research agenda? You do not need to arrive independent, but you should want to become independent.

Calibration #

Can you tell whether a project is too easy, too hard, too vague, or just right for the next six months?

Mathematical maturity #

Can you work with definitions, assumptions, examples, counterexamples, and proofs? You do not need to know everything. You do need to be willing to learn.

Experimental taste #

Can you build small simulations, prototypes, or empirical checks to test ideas quickly?

Writing #

Research exists socially through writing. Abstracts, proposals, papers, reviews, notes, and emails all matter.

Communication #

You should be able to explain what you are doing in one minute, ten minutes, and one hour. These are different skills.

Balance #

Good research is intense, but it should not require destroying the rest of your life. Sleep, exercise, friendships, and sunlight are not optional implementation details.

UC students #

MSc students at UC #

I am open to supervising MSc theses connected to my research areas. Good MSc projects should have a clear technical object and a realistic scope.

Possible MSc project types:

theory-guided simulation study;
implementation and evaluation of a probabilistic method;
small but rigorous extension of an existing model;
benchmark or dataset construction with a strong methodological angle;
AI-for-entertainment or co-creativity project with a serious technical core.

When contacting me, include:

your programme;
your background;
courses you enjoyed;
technical skills;
a short description of the kind of thesis you want;
whether you prefer theory, experiments, systems, or applications.

PhD students at UC #

If you want to do a PhD with me at UC/CISUC, the main route will often involve applying for an FCT PhD studentship or another external funding source.

A good PhD proposal should have:

a clear research question;
a plausible methodology;
a connection to my research areas;
a realistic 3–4 year plan;
evidence that you can write and work independently.

Please contact me several months before funding deadlines. “The deadline is next week” is usually not a research plan; it is a cry for help.

Portuguese, EU, and international candidates #

FCT PhD studentships #

For PhD candidates in Portugal, the standard route is often an FCT PhD studentship.

If you want to apply with me as a supervisor or co-supervisor, send:

CV;
transcripts;
1–2 page research idea;
writing sample;
relevant code/project links;
proposed PhD programme;
funding deadline.

A strong FCT proposal usually requires several iterations. I am happy to help strong candidates refine a proposal, but I cannot write the proposal for you.

FCT CEEC and other postdoctoral routes #

For postdoctoral researchers, FCT CEEC and related scientific employment calls may be relevant. These are competitive routes for candidates who already have a strong profile and an independent research direction.

If you want to discuss this route, send:

CV;
publication list;
1–2 page research programme;
why UC/CISUC is the right host environment;
what collaborations or training you want to develop.

MSCA Postdoctoral Fellowships #

MSCA Postdoctoral Fellowships can be an excellent route for postdocs who want to come to Coimbra with a strong research and training plan.

A good MSCA proposal is not simply “I will continue my PhD in Portugal.” It needs a research project, a training plan, a mobility logic, a career-development story, dissemination, communication, and a clear host–candidate fit.

If you want to apply for MSCA with me as host, contact me early, ideally 6–9 months before the deadline.

Please include:

date of PhD award;
countries where you lived/worked/studied in the last 36 months;
CV;
1-page project idea;
why UC/Coimbra;
what training you need from me and my network;
draft career-development narrative.

Brazilian candidates #

CAPES PDSE / doutorado sanduíche #

Brazilian PhD students interested in a doutorado sanduíche period can consider CAPES PDSE or similar institutional programmes.

I can support a strong application with an invitation letter and a research plan, but the candidate must coordinate with their Brazilian supervisor and graduate programme.

The best visits are not academic tourism. They should produce at least one of:

paper submission;
thesis chapter;
software artifact;
technical report;
benchmark/dataset;
mathematical result;
serious collaboration that continues after the visit.

When contacting me, include:

your Brazilian university and graduate programme;
your current PhD topic;
your supervisor’s name;
whether your supervisor supports the visit;
possible visit dates;
funding call;
1–2 page research plan;
expected output from the visit.

CAPES PrInt, Global.edu, and institutional routes #

Some Brazilian universities have internationalization routes such as CAPES PrInt, Global.edu, or local mobility programmes.

If your university has such a route, send me:

the call text;
eligibility rules;
deadline;
funding duration;
what documents you need from me;
a concrete research plan.

I am happy to look at concrete opportunities. I am less good at guessing secret bureaucracy from three continents away.

Brazilian students interested in FCT or MSCA #

Brazilian students may also be eligible for European routes depending on the call and their situation. For example, FCT PhD calls may be open to foreign citizens under specific conditions, and MSCA fellowships depend strongly on mobility and postdoctoral eligibility rules.

Always check the official call text. Funding rules change.

Visiting students and research stays #

I am open to research visits when there is a clear fit and a concrete plan.

A good visit has:

a defined research question;
a feasible output;
a start and end date;
funding or a clear funding application;
agreement with the home supervisor, when applicable;
a plan for integration with UC/CISUC activities.

Possible visit outputs:

joint paper;
thesis chapter;
tutorial/notebook;
software package;
benchmark;
workshop/summer-school material;
formal research note.

Example project ideas #

MSc-level projects #

Prior predictive checks for Bayesian neural networks.
Small GFlowNet implementation for structured inference.
Probabilistic AI for co-creative game design.
Simulation-based evaluation of recommender systems.
Tensor factorization and rank selection via moments.
Machine unlearning diagnostics for small language models.
Agent-based social simulation for entertainment AI.

PhD-level projects #

GFlowNets as amortized inference engines for probabilistic programming.
Prior predictive workflow for Bayesian latent-variable models.
Probabilistic methods for continual machine unlearning.
Bayesian tensor factorization and algebraic identifiability.
Constrained sampling and decision-making under uncertainty.
Probabilistic AI for creative, interactive, and entertainment systems.
Evaluation of multilingual mathematical reasoning in LLMs.

Postdoc-level projects #

Modern Probabilistic AI Workflow for Foundation Models.
GFlowNets as general-purpose amortized samplers.
Prior predictive design of Bayesian neural architectures.
Probabilistic machine unlearning and continual deletion.
European infrastructure for probabilistic AI workflow.
Bayesian and causal methods for adaptive AI systems.

What I can offer #

I can offer:

regular research meetings;
help turning fuzzy ideas into mathematical objects;
proposal feedback for strong candidates;
experience with top-venue ML writing;
connections to collaborators in Portugal, Brazil, Norway, Finland, Spain, and broader probabilistic-AI networks;
a friendly but rigorous environment;
too many diagrams, some of which eventually become notation.

I collaborate broadly across probabilistic AI, Bayesian modelling, GFlowNets, causal inference, machine learning, and computational creativity. A good student or postdoc can usually be connected to a wider research ecosystem rather than being isolated in a one-person supervision bubble.

What I cannot offer right now #

I cannot currently offer:

guaranteed salary without an external fellowship or project;
last-minute proposal miracles;
supervision for topics far outside my expertise;
a fully pre-packaged thesis where you simply execute tasks;
infinite meetings, despite my unfortunate tendency to be curious about everything.

How to contact me #

Use a subject line like:

Prospective [MSc / PhD / PDSE / MSCA / FCT / Postdoc] candidate — [topic]

In the email, include:

who you are and your current institution;
which route you are considering: MSc, FCT PhD, MSCA, CAPES PDSE, CEEC, research visit, etc.;
your deadline;
1–2 paragraphs on your research idea;
why my work is a good fit;
CV;
transcript or publication list, if relevant;
one writing sample;
GitHub/project page, if relevant.

I am more likely to answer messages that show a concrete research fit than messages generated by the sacred autocomplete spirits.

Before emailing: checklist #

Before emailing, please check:
Did you read at least one of my recent papers or project descriptions?
Can you name the research area you want to work in?
Do you know which funding route you are considering?
Do you have a rough deadline?
Do you have a CV ready?
Do you have a writing sample?
Can you explain why Coimbra/UC/CISUC makes sense for this plan?
Can you explain what you want to learn from me?

Frequently asked questions #

Do I need to already know advanced probability or Bayesian statistics? #

No. But you should be willing to learn mathematical tools when the research problem demands them.

Do I need to publish before contacting you? #

For MSc students, no. For PhD candidates, it helps but is not mandatory. For postdocs, a strong publication/writing record is important.

Can I work remotely? #

For informal collaboration, sometimes. For formal supervision, fellowships, or visits, this depends on the programme rules.

Can you fund me? #

Not usually at the moment. Most candidates will need to apply through FCT, MSCA, CAPES, or another funding route.

Can you help me write a proposal? #

I can help strong candidates improve a proposal when there is a clear research fit and enough time. I cannot rescue last-minute applications.

What makes a candidate stand out? #

A clear question, evidence of initiative, honest self-assessment, good writing, and the ability to move between ideas and implementation.

Disclaimer #

Funding calls, eligibility rules, deadlines, and documents change often. This page is only a guide. Always check the official call text and institutional instructions before applying.