Deep Learning Intern — Perception & Signal

About the Role

You will own Eunoie's deep learning model stack — from multimodal emotion detection to the signal processing pipelines that feed the emotional graph. You will work at the intersection of research and production, training models that run in real-time alongside live user conversations.

What you'll do

• ›Design and train multimodal emotion recognition models (text, tone, behavioural signals)
• ›Build and maintain training pipelines using PyTorch — data ingestion, augmentation, validation
• ›Deploy models to production inference endpoints with latency budgets under 100ms
• ›Instrument model performance with continuous evaluation against emotional benchmark datasets
• ›Collaborate with the Gen AI team to integrate classifiers into the generative loop
• ›Research and implement state-of-the-art approaches from affective computing literature

What we're looking for

• ›Strong PyTorch fundamentals — you write custom training loops, not just fit() calls
• ›Experience with sequence models: Transformers, LSTMs, or attention-based architectures for NLP
• ›Familiarity with multimodal learning (text + audio or text + behavioural data is a plus)
• ›Experience deploying models to production — not just notebooks
• ›Knowledge of model evaluation beyond accuracy: confusion matrices, calibration, fairness audits
• ›Prior work in affective computing, sentiment analysis, or emotion recognition preferred