Free guides
Questions, systematic reviews, proposals, paper structure, journal choice, and machine-learning experiments — each written to be used, not skimmed.
3 min read
Most cold emails to supervisors are ignored. Here is how to research, write, and send one that gets a reply, with a short template you can adapt.
Read the guide4 min read
A statement of purpose is a research narrative, not an autobiography. Here is a five-part structure, what to put in each part, and the mistakes that get applications ignored.
Read the guide3 min read
Turn a topic into a research question you can answer: the thing you will study, the comparison, the setting, and the result that would change your mind.
Read the guide4 min read
Where to look for gaps in the literature, the different kinds of gap, and how to check that yours is genuine and worth a thesis, before you spend a year on it.
Read the guide5 min read
A section-by-section outline for a PhD or small-grant proposal: problem, aims, what is already known, method, timeline, and risks, and what reviewers actually score.
Read the guide3 min read
Choose the method from the question, not from the tools you already know: experiment, observational study, qualitative study, or a review, and say what would count as an answer.
Read the guide4 min read
Plan a systematic review before you search: the question, the eligibility rules, the databases, the screening, and the synthesis. The flow diagram has its own guide.
Read the guide3 min read
What each box of the PRISMA 2020 flow diagram means, how to count records at every stage, and the arithmetic checks that catch mistakes before a reviewer does.
Read the guide3 min read
A working order for an AI research project: question, data, baseline, experiment, evaluation, and the paper. Skip a step and the GPU time usually does not come back.
Read the guide3 min read
Fix the dataset, the split, the baseline, the metric, and the number of runs before you train, so the result is something a reviewer can check.
Read the guide3 min read
Choose the split, the metric, and the baseline before you report a number. A single seed, a leaked test set, or the wrong metric will not survive a careful reader.
Read the guide3 min read
The mistakes that get an AI paper rejected, or published and then disbelieved: leakage, a weak baseline, one seed, metric shopping, and a claim the experiment never tested.
Read the guide3 min read
Which statistical test fits which machine-learning comparison: two models on one test set, several seeds, or many datasets. Includes a working paired-bootstrap example.
Read the guide2 min read
Three working SLURM job scripts for ML research: one GPU, a job array across seeds, and multi-GPU with torchrun, plus the mistakes that waste the most time.
Read the guide4 min read
Order a computer science or AI paper so a reviewer can find the contribution, the comparison, and the limit of the claim without hunting for them.
Read the guide3 min read
Shortlist a journal from its scope, its readers, where it is indexed, and the papers it published last year. Drop any venue you cannot verify.
Read the guide3 min read
A revision decision is not a rejection. Learn how to structure a point-by-point response letter that answers every comment, disagrees politely, and makes the editor's job easy.
Read the guide3 min read
Conference rebuttals are short, rushed, and read by tired people. Here is how to prioritise, what evidence to include, and the mistakes that waste your limited space.
Read the guide4 min read
A practical checklist for judging whether a journal is legitimate: the warning signs, the independent checks to run, and why no single list can do the work for you.
Read the guide3 min read
What examiners are really looking for, a week-by-week preparation plan, the questions to rehearse, and how to handle the ones you cannot answer.
Read the guide