Concentrated expertise
Teaching experience often lives in the writing and spoken knowledge of a few people.
Case study · Professor agent
Operational · open for pilotsA teacher that answers from what it knows.
An AI teaching agent built for francophone universities in Africa. It makes a corpus of texts and spoken knowledge searchable without inventing beyond its sources.
Case study · Professor agent
Make knowledge accessible without distorting it.
Recurring questions
Students and teaching staff need a way to retrieve that knowledge.
An honest limit
The agent does not fill gaps with invented answers.
Case study · Professor agent
How the agent works
Texts and spoken knowledge form the teaching corpus.
Illustrative workflow. Select a step to explore it.
See system details
Case study · Professor agent
Speak or upload text. Then ask.
01 / Share
An educator can contribute knowledge through speech or text.
02 / Retrieve
Retrieval finds relevant passages in the corpus.
03 / Answer
The answer is grounded in retrieved sources. An out-of-scope question is refused.
French first. Next.js · TypeScript · language model · local vector retrieval · speech-to-text.
Case study · Professor agent
Know how to answer. Know when to stop.
Illustrative example using a fictional passage, not actual corpus content. Passage: “The method presented has two steps: observe, then compare.”
Question: What are the two steps described?
Agent: According to the supplied passage, observe, then compare.
Question: What other method does the author recommend?
Agent: The supplied corpus does not contain that information. I cannot answer from my sources.
Case study · Professor agent
A tested scope, specific results.
8/8
Retrieval selected the right material for all 8 questions in the test set.
3/3
All 3 out-of-scope questions were correctly refused.
French relevance
Relevant passages score clearly above unrelated material.
September 2026 tests, per the supplied brief. This small test set does not measure universal accuracy.
Case study · Professor agent
Operational. Open for pilots.
A working system
The corpus can be queried with a clear boundary around its sources.
A next step
Explore a pilot in francophone education.
A transparent status
No institutional deployment, student count or partnership is announced.
Case study · Professor agent
Your business could be the next case.
Let’s explore a system built around your team’s work.
Lusona Autonomous, Montréal