EdTech Provider · EdTech · Higher Ed
A fine-tuned classroom assistant — trained on the course, not the internet.
A classroom AI agent fine-tuned on course materials that answers student questions instantly, distributes resources, and supports faculty triage.
2,400+
student questions answered weekly
−65%
faculty time spent on Q&A triage
4.8 / 5
student satisfaction with answer quality
EdTech · Higher Ed
EdTech Provider
Client
EdTech Provider
EdTech · Higher Ed
Headline metric
2,400+
student questions answered weekly
Deliverables
4
shipped to production
Stack
4+ tools
across the build
Challenge
Faculty were drowning in repetitive student questions about course content, deadlines, and resources. Generic LLM tools hallucinated on course-specific material. The provider needed an agent that was reliably accurate on the curriculum.
Approach
Fine-tuned an open-weight Llama 3.1 model on the curriculum corpus, supplemented with RAG over course documents for current information. Built a refusal layer when the model wasn't confident, plus a faculty review queue for unclear questions.
Outcome
Over 2,400 student questions answered weekly with 4.8/5 satisfaction. Faculty triage time dropped 65%, freeing them for higher-value teaching. The agent has been the platform's top-engaged feature for six consecutive months.
Over 2,400 student questions answered weekly with 4. The team owned this end-to-end.
EdTech Provider
EdTech · Higher Ed
student questions answered weekly
faculty time spent on Q&A triage
student satisfaction with answer quality
What we shipped.
- Fine-tuned classroom agent
- Course-document RAG
- Faculty review console
- Analytics dashboard
The tools we used.
More case studies
All case studies
EdTech · Series A
Mockinto
47%
An AI interviewer that actually adapts to the candidate.
A multi-turn interview agent that adjusts difficulty in real time, scores answers against a rubric, and gives structured feedback in under two seconds.
Read case
Healthcare · Compliance
Healthcare Information Platform
+22%
A fine-tuned Llama 3.1 model for healthcare inquiries.
Domain fine-tuning that beat general-purpose models on accuracy, latency, and cost — while keeping data inside the customer's VPC.
Read caseLet's talk about your version of this.
30-minute discovery call. No NDA gatekeeping. We'll tell you straight whether we're a fit.
info@aiinfox.com