We build AI into business software where it has a clear job: assistants that answer, models that forecast from your own data, and systems that read text and images. Where ordinary software would do the job better, we say so. The assistant on this page is live, so ask it about Durrani Lab, or describe an idea and it will draft a brief for our team.
What we build
Each of these is built into software a business already runs, or into a new system we build around it.
- 01
AI assistants and integration
An assistant inside your product or on your site that answers from your own content and hands over to a person when it should.
- 02
Natural language processing
Software that reads, sorts and summarises text: enquiries, documents, support tickets and reviews.
- 03
Computer vision
Software that reads images and video, for example scanned documents, products on a shelf or defects on a line.
- 04
Machine learning models
A model trained on your records to answer one question, and tested against results you already know.
- 05
Predictive analytics
Forecasts drawn from your own history, such as demand, repayments or stock.
- 06
Data analytics
Reports and dashboards that put the figures a team needs in one place, kept up to date.
Plain advice
Part of the work is saying no. These are the cases where we recommend ordinary software instead.
- 01
The rule is already known
If the answer follows from a rule you can write down, we write the rule. It costs less, runs faster and gives the same answer every time.
- 02
The answer must be exact
A model can be wrong and sound sure. Sums, balances and legal wording belong in ordinary software, with AI kept to the parts where a draft is useful.
- 03
There is no history to learn from
A forecasting model learns from records. Where there are none yet, we say so and suggest what to start collecting.
- 04
The running cost outweighs the saving
Every answer from a large model costs money. We estimate that cost before building and set it beside what the feature saves.
How the work runs
Which data the system may use, where it is processed and who can see it are settled with you before any of it is touched.
- Step 01
Assess
We look at the problem and the data you hold, and say plainly whether AI suits it.
- Step 02
Prototype
A small working version on a sample of real data, so the idea is judged on results.
- Step 03
Build
The feature is built into your software with limits, logging and a fallback for when the model is unsure or unavailable.
- Step 04
Run
After launch we watch quality and cost, and adjust the instructions, the model or the data as they change.
The assistant on this site
The assistant answers questions about Durrani Lab: what we build, how a project starts and how we arrive at a price. Describe what you want built and it will draft a project brief, then send it to our team once you agree.
- 01
It is an AI
You are talking to a language model, not a person. It can make mistakes, and it will not quote prices or dates.
- 02
What happens to what you type
Your messages go to an AI model provider to produce each reply. This site does not store the conversation, and it is gone when you reload the page.
- 03
When a brief is sent
A brief is emailed to our team only after you give a name and an email address and say it may be sent.
How we handle your details is set out in the privacy policy.
Questions
What AI and machine learning work do you do?
We work on predictive analytics, natural language processing, computer vision, machine learning models, data analytics, and AI integration into existing products.
What is the assistant on this page?
It is a live AI assistant that answers questions about Durrani Lab. It can also turn your idea into a project brief and send that brief to our team. If the assistant is unavailable, the contact form and WhatsApp on +91 82 4010 4020 work in its place.
Can you add AI to software we already have?
Yes. AI integration is one of our services. We look at the product and the data it holds, and identify where a model would be useful before any work is proposed.
Do we need a large amount of data to use AI?
It depends on the problem. Some features work with existing language models and little data of your own, while predictive models need historical records. We review your data first and tell you what is realistic.
How much does an AI project cost?
The cost depends on the problem, the state of your data, and the running cost of the models used. We quote after an assessment, and the quote separates the build cost from the ongoing running cost.
Related services
CRM, ERP and internal systems for organisations that have outgrown their spreadsheets.
More on custom softwareCloud architecture, migration and release pipelines for teams whose software has to stay up.
More on cloud and DevOps