AI engineering for a defined problem in your product.

Document AI

Extract fields from documents, including mixed Bengali and English.

We combine OCR and LLMs. Scoping starts with three questions: which documents you need to process, which fields matter, and what the receiving system expects.

Computer vision

Detection, segmentation and tracking.

We detect objects, segment images and track objects in video. The pilot focuses on the task and footage that matter to your application.

Assistants and agents

Connect an LLM to the tools your team uses.

We build assistants and agents connected to your tools. The scope sets out what information they can use and which actions they may take.

Voice AI

Voice AI for customer support and sales enquiries.

Start with a voice AI pilot for one support or sales workflow. Define what it may answer, which details it should collect and when a person should take over.

A support pilot can cover routine questions and ticket creation. A sales pilot can capture requirements, qualify an inbound enquiry or request an appointment through your existing tools. We agree the permitted actions and test the handover and failure cases before live use.

You receive the custom code and configuration in your repository, the evaluation results and the known limitations. Deployment, usage and ongoing support are scoped separately.

English is the starting scope. Bengali and mixed-language workflows can be assessed in a separate language test.

Evaluation sets

Decide what the system will be judged against.

We build evaluation sets for the task. During scoping we agree the expected outputs, the errors that matter, and the cases that should go to a person for review.

Pilot and delivery

Agree the first test before the larger build.

We sign an NDA before reviewing your data. The paid pilot covers one problem, its inputs and how the result will be assessed. You see a demo every sprint, and the code stays in your repository. Further scope follows the pilot.

Pilot pricing depends on the data, the integrations needed, the evaluation work and the handover. We quote after scoping.

Handover

What your team gets to take the work forward.

• Code in your repository
• Setup instructions
• The agreed evaluation material and results
• A record of known limitations

Deployment, ongoing support and later changes are agreed separately.