Responsibilities :
- Build and evaluate prototypes / POCs for generative AI features and ideas.
- Fine-tune and adapt open-source LLMs and smaller generative models for specific use cases.
- Collaborate on multimodal experiments involving text, image, and audio data.
- Implement data preprocessing, augmentation, and basic feature engineering for model inputs.
- Design and run experiments, define evaluation metrics, perform ablations, log results, and iterate on models.
- Optimize inference and memory footprint using techniques such as quantization, batching, and basic distillation.
- Contribute to model training pipelines, scripting, and reproducible experiments.
- Work with cross-functional teams (product, infrastructure, MLOps) to prepare prototypes for deployment.
- Write clear documentation, present technical results, participate in code reviews, and share knowledge with the team.
- Continuously learn by exploring research papers, new tools, and innovative approaches in AI.
Mandatory Technical Skills :
Strong Python programming and familiarity with ML tooling (numpy, pandas, scikit-learn)2+ years hands-on experience with PyTorch and / or TensorFlow for model development and fine-tuningSolid grounding in classical ML and deep learning : supervised / unsupervised learning, optimization, CNNs, RNNs / LSTMs, TransformersGood understanding of algorithms, data structures, numerical stability, and computational complexityPractical experience fine-tuning open models (Hugging Face Transformers, LLaMA family, BLOOM, Mistral, or similar)Familiarity with PEFT approaches (LoRA, adapters, QLoRA basics) and efficiency techniques (mixed precision, model quantization)Comfortable running experiments, logging results (Weights & Biases, MLflow), and reproducing experimentsExposure to at least one cloud ML environment (GCP Vertex AI, AWS SageMaker, Azure AI) for training or deploymentEffective communication skills for documentation and cross-team collaborationSkills Required
Python, Pytorch, Tensorflow, Aws