← Back to projects
// case study

Reprebot — Fine-tuned AI for UNAL

Qwen 3.5 (4B) fine-tune for UNAL with its own API, deployed on WhatsApp, CEIS chat and forum.

Screenshot of Reprebot — Fine-tuned AI for UNAL

01 · The problem

Context

Dozens of repetitive queries about academic regulations and procedures arrived daily via WhatsApp and forums without the human capacity to respond in real time.

02 · What I did

Implementation

A Qwen 3.5 (4B) fine-tune on UNAL regulations with a RAG engine over official sources, a dedicated API with streaming, and multi-channel deployment: WhatsApp, CEIS chat and forum, with answers that cite their sources.

03 · Result

Status

Live chat at https://repre.nikko.dev and production API, answering regulations with cited sources across all three CEIS channels.

04 · Stack

Tools

Qwen 3.5 4BFine-tune UNALFastAPIRAGWhatsAppAPI