Revelation Software Labs · Bengaluru
Practical AI. Built for real work.
We build software that helps people get things done. Our current research explores how Kannada conversations can become clear, dependable business requests.
- Founded
- 2018
- Based in
- Bengaluru
- Funding
- Bootstrapped
- Current focus
- Kannada AI research
Current research · KAR
Kannada conversations. Clear next steps.
We are developing technology for assistants that ask the right questions, confirm important details and carry out authorised tasks. Our first test environment is a fictional college appointment desk.
-
For students
Ask in Kannada or English
Speak or type a question or a request, and get a plain answer or a clear summary of what was asked for.
-
For college staff
Stay in control
Staff decide what the assistant may answer, which requests come to them, and which tasks, if any, it may complete.
-
For integrators
Connect what is allowed
Where a college wants tasks completed automatically, the assistant connects to its existing systems with permission for each task.
Existing AI models can already understand speech and use software tools. Our research asks a narrower question: can a carefully designed assistant complete a college’s real tasks more dependably, and be worth running day to day? We are measuring that now. It is not yet shown.
Proposed college experience
What could this look like for a college?
This is the journey we are designing towards, so that a college never has to learn new technology to use it. It is not available yet.
-
Step 1
The college shares approved information
Courses, fees, documents, common questions, campuses and appointment rules, as the college wants them told.
-
Step 2
RSL sets up a student link
One Kannada and English assistant link the college can share on its website, on a notice-board QR code or in a message.
-
Step 3
Students ask
They speak or type, in Kannada or English. The assistant asks a short question when something is unclear.
-
Step 4
Staff review, or approved tasks complete
Requests go to the right staff member. Only tasks the college has authorised and connected are completed automatically.
Three kinds of help. The college switches each one on.
A first pilot would begin with the first two. Automatic tasks come later, and only with setup and integration.
-
Start here
Answer questions
Answers come only from the information the college has approved. If the answer is not there, the assistant says so and points to the right office.
Needs Approved information
-
Start here
Send requests to staff
The assistant collects the details, reads them back to the student and sends the request to the responsible staff member.
Submitted The student is told it is waiting for staff, not that it is booked.
Needs Staff contacts and rules
-
Later, after integration
Complete approved tasks
For tasks the college has authorised, such as moving an appointment, the assistant asks for a clear yes and then completes it through the college’s system.
Confirmed Said only after the college’s system has accepted it.
Needs Integration and permission
-
Student
I’d like a counselling appointment on Saturday at ten.
CheckReads the details back before doing anything.
-
Assistant
Counselling, Saturday at 10:00, North Campus. Shall I book it?
-
Student
Actually, can we make it Sunday?
ChangeSunday replaces Saturday. Nothing was booked for Saturday.
Replaced:
Saturday · 10:00 -
Assistant
Counselling, Sunday at 10:00, North Campus, instead of Saturday. Shall I book Sunday?
-
Student
Yes, please book Sunday.
Clear yesThe yes is for Sunday, the details on screen.
-
Assistant
Done. Your counselling appointment is confirmed for Sunday at 10:00.
Said only after the college’s booking system accepted it.
- Appointment
- Confirmed · Sunday · 10:00 · Counselling
- Saturday
- Never booked
For developers: how the research prototype handles this
In our fictional sandbox, each read-back is a proposal with its own identity. A yes applies only to the proposal the person was shown. A late or out-of-date yes is rejected, a correction withdraws the earlier proposal, a retried request does not create a second booking, and “done” is said only after the booking tool returns a receipt.
These behaviours are checked by automated tests on synthetic text cases. They have not been tested with real students, real speech or a real college system.
Proposed college experience
What a college would provide
RSL would help collect and set this up. The college would not need its own technical team to start.
Approved information
Courses, fees, required documents and answers to common questions.
Campuses and hours
Locations, offices and contact hours.
Appointment rules
Which appointments students can request, and when.
Responsible staff
Who receives each kind of request and who approves it.
Permissions and system access
Only if the college wants some tasks completed automatically.
Status · October 2026
What is available today
KAR is a research prototype. It is a working name, not a launched product or a registered brand.
Working in the lab
- A text sandbox of a fictional college appointment desk.
- Kannada speech input, built and waiting for live testing.
- Evaluation tools that replay synthetic test conversations.
Not yet
- No college is using it.
- No general answering of college questions.
- No staff dashboard.
- No public demo.
Company
Revelation Software Labs
We started Revelation Software Labs in Bengaluru in 2018 and have funded it ourselves from the first day. For most of that time we designed and shipped software for clients, and built our own products on the web and on blockchains.
Today our own work is in AI. RSL is the company, and it may build more than one product. KAR is its current research project.
- Founded
- 2018
- Bengaluru, Karnataka
- Funding
- Bootstrapped
- No outside capital raised.
- Background
- Client software
- Plus our own web and blockchain products.
- Now
- AI research
- Starting with KAR.