M.Sc. student in Image Analysis and Machine Learning at Uppsala University (Sep 2025 – Jun 2027), building on a B.Sc. in Data Science from NOVA IMS (Lisbon) and an Erasmus semester at Lund University. I work on computer vision and deep learning, and I'm currently a research intern at Furhat Robotics (Stockholm) on multimodal perception. Portuguese, based between Stockholm and Uppsala.
- Semi-supervised distillation for multimodal cancer-cell classification — on a hard, patient-disjoint 12-patient Kaggle benchmark, I grew the effective training set with hard pseudo-labels (Lee 2013) then soft-target knowledge distillation (Hinton 2015) on a dual EfficientNet-B0 with AdaBN — the change that mattered more than scaling the network. Full LaTeX methodology write-up and an honest failure analysis. → multimodal-microscopy-distillation
- Vision Transformers for emotion recognition — fine-tuned ViT-Base on RAF-DB for facial-emotion recognition, feeding the dialogue layer of a Furhat social-robot tutor (group project). → furhat-emotion-tutor
- Deep learning from scratch — a NumPy-only fully-connected network with hand-written backprop, then a PyTorch progression to a residual CNN on MNIST. → mnist-from-scratch-to-resnet
- Furhat social robot for student mental-health check-ins — a group HRI study (1MD043) testing whether empathetic vs. neutral robot behaviour changes how students perceive a wellbeing check-in (between-subjects, n = 23, RoPE + Godspeed scales); I contributed to the experimental design and analysis and added a Gemini LLM backend. → HRI_Furhat
- Multimodal mushroom-ID chatbot — a Gradio app that reads a photo, extracts a strict-JSON description, and answers as a mycology assistant with a safety guard against foraging advice; runs on Google Gemini or a Hugging Face pipeline (Llama 3.1 + BLIP captioning). → mushroom-id-chatbot
- LoRA fine-tuning of Qwen3-0.6B — a small-LLM SFT experiment on a BeautifulSoup-scraped corpus, with a Gradio base-vs-fine-tuned comparison. → greek-mythology-chatbot
Languages: Python (daily), SQL, R, MATLAB, Bash Deep learning & computer vision: PyTorch, Hugging Face Transformers, timm; Vision Transformers, CNNs, transfer learning, knowledge distillation, semi-supervised learning, PEFT / LoRA Also: scikit-learn, NumPy, Pandas; Gradio; Git, Jupyter, Kaggle; LLM APIs (OpenAI, Gemini)
Research intern at Furhat Robotics (Jun – Dec 2026). Looking for computer-vision, ML / AI engineering, and deep-learning research roles — and a master's thesis — in Sweden and the EU.
- Email: [email protected]
- LinkedIn: linkedin.com/in/rafael-alexandre-proenca