Location: Paris / Partial remote work possible
Contract type: Full-time permanent position (CDI)
Start date: As soon as possible

We are looking for an experienced AI researcher in medical speech analysis to join our team in Paris and help advance life-saving voice-based technologies.

About  e-sensia

At e-sensia, we develop cutting-edge medical assistance technology based on voice analysis.
Our mission: save lives by enabling faster and more accurate detection of medical emergencies using artificial intelligence.

Our solution is already used by major emergency response organizations to analyze real-time calls.
It is powered by a unique speech-processing pipeline, enriched with millions of real-world emergency calls and associated medical data.

Responsibilities

As an AI researcher in medical speech analysis, you will play a key role in advancing the scientific capabilities of our product by designing and developing innovative models, including:

• Fine-tuning of large audio language models (Whisper, Wav2Vec, EnCodec, etc.)
• Raw speech or spectrogram processing, optionally combined with video for multimodal approaches
• Automatic detection of critical events (breathing irregularities, distress, emotional states, silence)
• Recognition of subtle vocal patterns associated with medical conditions
• Close collaboration with clinicians to validate AI models in real-world emergency scenarios
• Contributing to the scientific strategy of the company
• Participation in scientific publications, patent filings, or collaborative European research projects

Profile

• PhD or equivalent experience in AI, signal processing, or deep learning for audio/video
• Proven ability to initiate and lead research projects
• Strong expertise with deep learning frameworks, including:
  – PyTorch for fine-grained modeling and experimental workflows
  – Hugging Face Transformers for rapid deployment and adaptation of pretrained models (Wav2Vec2, Whisper, Hubert, etc.)
  – TensorFlow/Keras or JAX/Flax as complementary tools for prototyping and benchmarking
• In-depth understanding of modern neural architectures for audio and video, including:
  – Self-supervised models: Wav2Vec 2.0, HuBERT, Whisper, EnCodec, AudioMAE
  – CNN and RNN architectures for long time-series modeling
  – Transformer-based models for audio (AST, conformers, SpeechTransformer)
  – Multimodal fusion techniques (audio + video) are a strong plus
• Hands-on experience with raw audio signal processing, including:
  – Feature extraction: MFCCs, log-Mel spectrograms, pitch, formants, etc.
  – Data augmentation: SpecAugment, noise injection, vocal tract length perturbation
  – Temporal/frequency analysis: FFT, STFT, CQT
  – (Bonus) Synchronized video analysis (e.g., lip reading, emotion recognition)
• Ability to optimize models for real-time deployment, including:
  – Quantization, pruning, knowledge distillation
  – Deployment to GPU environments using ONNX, TorchScript, or Triton
• Strong interest in healthcare and socially impactful technologies
• Professional English required; spoken French appreciated

Why join us?

• A mission-driven environment where every algorithm can help save lives
• A deeptech startup backed by major players in healthcare and innovation
• A scientifically rigorous and supportive work culture
• A passionate and multidisciplinary team working on real-world impact

How to apply

Please send your resume, a short motivation message, and if possible, an example of a relevant project or publication to:
job@e-sensia.com