The person behind the gradient.

I’m Afonso Oliveira — an AI and software professional with 14 years across research, product, and engineering. I build AI systems end-to-end: from the theory and design to the deployment, from the prompt to the infrastructure.

My recent work spans behavioral-data analysis, NLP classification, LLM agent development, and observability. Before that, I spent years in telecommunications research, applying neural networks to 5G resource allocation. The common thread: making AI do useful things in real systems, then measuring whether it actually did.

Background

I hold a PhD in Electrical and Computer Engineering from Instituto Superior Tecnico (University of Lisbon), where my thesis explored machine learning for network slicing in next-generation mobile networks. That led to 10 peer-reviewed publications and 150+ citations.

Professionally, I’ve worked across contexts that don’t usually overlap:

  • Startups — co-founded adCaffe, built early-stage products from scratch from different ideas, learned what ships, what doesn’t, and when to pivot. Mostly when to pivot.
  • R&D — eight years at INESC-ID, building simulators, writing papers, connecting things that weren’t designed to be connected
  • Enterprise — at Siemens, turned project-specific ML code into a reusable AI product, cutting proof-of-concept delivery from one week to under one hour
  • AI leadership — led AI strategy and execution for a mental health product, from classical ML to LLM agents with a set up observability stack

What I work with

AI & ML — classical ML, NLP, LLM agents, model evaluation and observability.

Engineering — Python is my primary language. Also comfortable in JavaScript/TypeScript, SQL, C/C++, and Flutter. I use FastAPI for backends, Terraform for infrastructure, and Docker in everything I can touch.

Cloud — GCP, Azure, AWS. I’ve deployed models and managed infrastructure across all three.

The full stack, when the situation calls for it. I’m a generalist by necessity — startups don’t have the luxury of specialists, and curiosity doesn’t respect job descriptions.

Publications

A selection — the ones that survived peer review and my own re-reading:

  • “Low-power and lossy networks under mobility: A survey” — Elsevier Computer Networks, 2016 (105 citations)
  • “Generating Synthetic Datasets for Mobile Wireless Networks with SUMO” — ACM MobiWac, 2021 (21 citations)
  • “Towards green machine learning for resource allocation in beyond 5G RAN slicing” — Computer Networks, 2023

Full list on Google Scholar.

I also contributed a chapter to “88 Vozes sobre Inteligencia Artificial”, published by ISCTE in 2023.

Elsewhere

Contact

You can reach me at [email protected].