One small language model, trained for one job your team repeats every day.

I fine-tune language models under one billion parameters on your company's own examples, small enough to run on a laptop or a modest server, and connect them to the workflow where the task happens. It starts with a consultation, and I only build if a small model is the right tool.

Where it fits

The model sits between the work your team already does and the tool it already uses. It learns from your examples, handles one task, and passes the result on.

Where a small model fits in your workflow A diagram showing three stages: Your examples on the left, connected to a Small model in the center, which connects to Your workflow on the right. Your examples Past work, done correctly Small model Under one billion parameters, trained on your examples Tested on examples it has not seen Your workflow Connected to one tool you already use Where a small model fits in your workflow A vertical diagram showing three stages: Your examples at the top, connected to a Small model in the center, which connects to Your workflow at the bottom. Your examples Past work, done correctly Small model Under one billion parameters, trained on your examples Tested on examples it has not seen Your workflow Connected to one tool you already use

Why small

Runs where your data already is

A model under one billion parameters can run on a laptop or a modest server, so your examples do not have to be sent to a third-party model provider each time it runs.

Built for one job

A model trained on your own examples of one task can be more consistent on that task than a general-purpose model prompted to do it. I test this on examples the model has not seen before handover, and the report shows what it gets right and wrong.

Cheap to keep running

Small models need little hardware, so running costs stay small, and there are no per-request fees from a model provider when the model runs on your own machine.

Tasks a small model can take on

Sorting incoming emails or support requests into the right category.

Pulling named fields, such as dates, amounts and names, out of invoices, forms and messages.

Drafting standard replies that a person reviews before sending.

Tagging and routing records in a spreadsheet or database.

Is it a fit

A good fit

The task repeats every week and a person does it by hand.

You have past examples of it done correctly, as emails, documents or spreadsheet rows.

A wrong answer can be caught by a person or a simple check.

Probably not a fit

Open-ended questions that need broad general knowledge.

Long documents that need careful reasoning across many pages.

Tasks where you have no past examples to learn from.

A two-week pilot

Scope
One task, one model, one integration.
Timeline
Two weeks from the day your examples arrive.
Price
Fixed, quoted in writing after the first call.
Not included
General-purpose chatbots. Tasks that need open-ended reasoning or long documents. Hosting and maintenance after handover unless agreed separately. Any accuracy promise beyond what the test report shows.

How it works

  1. Quote: You send a request. I reply within two working days with questions or a fixed quote.
  2. Day 0: A 30-minute call to confirm one repetitive task worth automating. If a small model is the wrong tool, I will tell you.
  3. Days 1 to 3: We agree how success is measured, and I prepare your examples for training.
  4. Days 4 to 10: I fine-tune the model and test it on examples it has not seen.
  5. Days 11 to 14: I hand over the model, a plain-language accuracy report, and instructions to run it.

Questions

Do I need to be technical?
No. I need examples of the task and a description of where it happens. I handle the training and connect the result to one tool you already use.
Where does my data go?
Your examples are used to train the model. I will agree in writing how they are handled, stored and deleted before you send anything.
How accurate will it be?
I will not promise a number in advance. After training I test the model on examples it has not seen and give you a plain report of what it gets right and wrong, so you can decide whether to use it.
What if a small model is the wrong tool?
Then I will tell you on the first call, and you will have spent 30 minutes, not two weeks.
What happens after the two weeks?
You receive the model, the report and instructions to run it. Hosting and maintenance are separate, and only if you want them.

About

I'm Prosun Banerjee, founder and CEO of Brevix. I have built machine learning projects and workflow automation, including during internships. I work on my own, so you speak directly to the person who builds your model.

Tell me which task your team repeats most.

Request a quote