DSTI TechBlog: ideas, systems and projects worth examining.
Faculty research, student-built datasets, alumni work in the field, and engineering decisions from inside DSTI School of Engineering. Each article begins with something real: a model, a paper, a repository, a system or a difficult trade-off.
When the data does not exist
Why DSTI teaches agent-based modelling when the dataset is absent, fragmented or rightly out of reach.
Read the long-form articleBuilderLab: from solo learning to a shared engineering workflow
A live beta examining team formation, state modelling, access control, maintenance and the evidence behind a shared software project.
Read the project accountEfficient IT begins before the code
Why the BSc follows energy, climate and computing across three years before asking what efficient IT really means.
Read the articleWhy we teach the history of computer systems in the BSc
Why computing, AI, networks, data systems and connected objects need historical grounding, technical memory and social context.
Read the articleWhen automotive market analysis becomes a data-engineering problem
A Google Cloud workflow connecting vehicle sales, powertrain components, suppliers, pricing and market addressability.
Read the project accountFrom complex drug databases to reproducible R workflows
How dbparser turns DrugBank, OnSIDES and TWOSIDES into consistent, traceable R objects for reproducible research.
Read the articleWhen a link carries meaning
Knowledge graphs, Linked Data and open standards: how explicit meaning changes the architecture of AI systems.
Read the articleWhen the model already knows the physics
Why intelligent systems should combine observations, optimisation and physical models instead of relearning nature from scratch.
Read the articleBuilding IdiomX: language beyond the literal
A reproducible multilingual dataset and four-task benchmark for detecting, retrieving and interpreting idioms across English, Arabic and French.
Explore the projectPredicting mental health risk through social media
A first-place approach combining LLM post classification with lightweight temporal modelling to forecast an unseen future risk level.
Read the research accountScaling a random forest without losing the statistics
Why scaling Random Forest models can change the statistics, and how out-of-bag error and variable importance became VSURF.
Read the articleRebuilding dsti.school: eighteen months on the road, three weeks at the keyboard
How international field insight became a deterministic, multilingual and AI-portable publishing system without losing the human reasons for building it.
Read the engineering accountHave a technical story worth documenting?
Students, alumni, faculty and Team DSTI can propose an account grounded in real work, evidence and lessons learned.