01Language
Reads and writes text. Coding, clerical work, documents, mail, chat, reports.
Same stack for a clerk or a programmer. Drafts, edits, files, and answers in the house style.
02Vision
Looks at pictures and says what’s in them.
Photos, screenshots, diagrams, UI. What is on the page, what is wrong with it, what to do next.
03Voice
Talks and listens. Calls, transcription, agents that speak.
Inbound and outbound. Live or recorded. Notes, scores, and handoff when a person should take over.
04Image and video
Makes and serves stills and motion.
Generation, variation, and delivery. Still frames or clips, on demand or in a pipeline.
05OCR
Pulls the words off scans, PDFs, photos, and forms.
Keeps the layout when it matters. Tables, handwriting, stamps, bad copies.
06Classifiers
Puts things in buckets. Yes or no, risk, category, spam.
Labels at the volume the rest of the shop already runs. Scores when a hard cut is not enough.
07Fine-tunes
Takes a general model and trains it on your work until it behaves like your shop.
Your documents, your calls, your labels. Smaller, cheaper, and less inclined to wander.
08Training
Builds a model when nothing on the shelf is enough.
Data, runs, evals, and the weights at the end. Used when a fine-tune will not cover the job.
09Serving
Keeps the models running so the product stays up.
Inference, queues, limits, and failover. The part that has to work on a Tuesday.