Fine-Tune a Transformer for Customer-Support Triage at an Enterprise AI Vendor
Overview
What this challenge is about.
You receive 240,000 labeled support tickets across 14 queues, with English, Bahasa Indonesia, and Tagalog. Fine-tune a multilingual transformer encoder (XLM-RoBERTa-base is a strong starting point). Evaluate macro-F1, per-language F1, and per-queue confusion. Compare against the TF-IDF logistic regression baseline. Package the model for inference (FastAPI + ONNX or TorchScript) and document p95 latency at production-realistic batch sizes.
The Brief
What you'll do, and what you'll demonstrate.
Cut misrouting error in half versus the TF-IDF baseline using a fine-tuned multilingual transformer, with production-grade inference.
Earning criteria — what you'll demonstrate
- Fine-tune transformer encoders on multilingual classification
- Evaluate multilingual classifiers per language and per class
- Export and serve transformer models in production
- Translate model improvements into SLA-relevant business framing
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
NLP Engineer
Multilingual transformer fine-tuning with production deployment is the canonical NLP engineer deliverable at enterprise-AI vendors.
This challenge sharpens
- transformers
- fine-tuning
- multilingual-nlp
Machine Learning Engineer
Owning the end-to-end fine-tune-to-deployed-service pipeline is exactly junior MLE territory at customer-support-automation companies.
This challenge sharpens
- fine-tuning
- inference-deployment
- pytorch
MLOps Engineer
ONNX export, quantization, and latency profiling at production batch sizes is core MLOps responsibility on model-serving teams.
This challenge sharpens
- inference-deployment
- pytorch
- fine-tuning