Learn the work behind Data, AI & Forward Deployed Engineering
Practical explanations, career decisions and reproducible workflows. Read the reasoning, inspect the evidence and follow the next skill into a real programme.
Text embeddings: evaluate similarity against labelled pairs
An embedding maps text to a vector. Cosine similarity then measures angle, not task relevance. Before using it for deduplication, retrieval or clustering, evaluate labelled pairs that represent what “similar” means for t
Text features: control vocabulary and fit boundaries
TF-IDF learns a vocabulary and document-frequency weights. Fitting on all documents before a split imports test-language statistics into the representation, even when no target is used.
Threat-model an AI application from data flows
A list of AI risks becomes actionable only when it is tied to the application’s assets, identities, boundaries and operations. Draw the real data flows before selecting mitigations.
Threat-model tool misuse and data exfiltration paths
An agent connects untrusted language to capabilities. Threat modelling must follow the complete path from user and retrieved content through model output, tool arguments, credentials, network destinations and stored resu
Threshold selection with unequal error costs
A single false-negative cost assumes every missed event has the same consequence. Many decisions are heterogeneous: missing a high-amount transaction may matter more than missing a small one, while review consumes a roug
Time-aware validation for a changing business process
A time-aware evaluation should recreate what could have been known when a model was fitted and used. Sorting by prediction date is necessary for many forecasting settings, but it does not ensure that the training labels
Time-series train-test splits: reproduce a real forecast date
A time-series split should reproduce a decision that could have happened on a named date. Randomly shuffling monthly demand mixes later conditions into earlier training data. The resulting score answers how well a model
Tokenization choices that change a text model
Tokenization decides which patterns a text model can represent. A word tokenizer may split sign-in into sign and in, while character n-grams preserve fragments across punctuation and spelling variants. The choice changes
Tokens and context windows: budget a real document task
A context window is shared space. System instructions, conversation history, retrieved evidence, tool schemas and generated output all consume it. Sending a document that fits by itself can still overflow once the applic
Tool errors: distinguish retryable and permanent failures
Retrying every exception wastes time and can repeat writes. Error contracts should distinguish transient availability from invalid requests, denied authority and business rejection.
Topic modelling: validate topics with human inspection
Topic models produce weighted term patterns. A low reconstruction error or coherent-looking word list does not establish a useful business theme. Reviewers must inspect terms and representative documents, name the theme
Trace an LLM request without logging private content
An observability system should answer which stage failed, which release was involved and how long it took. It does not automatically need the user’s entire prompt, the generated response, an email address or a bearer tok
Not sure which programme fits?
Tell us your background and we will map it to the right entry point — including saying so when a cheaper programme is the better fit. A counsellor replies within one working day.