Training a language model
PyTorch · Transformer modelsDesigned, trained and evaluated a 750 million parameter transformer on 5 billion tokens, with checkpointing and fault-tolerant training pipelines on A100 GPUs.
AI engineering & applied mathematics
Based in the Netherlands
Dewi Batista.
From mathematical ideas
to AI people can use.
I’m an AI engineer and consultant at Artefact. I build language model systems, generative AI tools and the pipelines behind them.
Selected work
Research, engineering, practice
Google × VELUX · WorkshopAI for consulting decisions
LLM applications · ArtefactBuilt an internal pricing assistant used by directors and managers to check project scope and support statement-of-work negotiations.
Generative models for forensics
Probabilistic modelling · ConsultancyAdvised the Netherlands Forensic Institute on generative probabilistic models trained on forensic data, supporting the interpretation of evidence in forensic reporting.
A little about me
Mathematics
meets making.
My background is in applied mathematics, machine learning and statistics. I enjoy the work between understanding a model and building something useful with it.
At Artefact, I work on AI systems for clients and internal teams. Before that, I taught neural networks, machine learning and cryptography at the University of Groningen, and automated reporting as a data engineer in Wales.
AI Engineer & Consultant
Artefact · Netherlands
Teaching Assistant
University of Groningen
Data Engineer
Coleg y Cymoedd · Wales
MSc Applied Mathematics
Machine Learning & Statistics · University of Groningen
Python / PyTorch / SQL / GCP / FastAPI / Docker