ML Engineer

Описание

We are seeking a highly motivated and skilled Machine Learning Engineer to join our team

focused on training and deploying machine learning models and fine-tuning large language

models (LLMs). This role will be central to developing efficient, scalable, and domain-specific

ML/LLM solutions that power our AI products.

Responsibilities:

• Design, implement, and optimize training pipelines for small-scale machine learning models

(e.g., decision trees, gradient boosting, small neural nets, CNN).

• Fine-tune large pre-trained language models (e.g., LLaMA, Mistral, Qwen, GPT) for specific

tasks using supervised learning.

• Conduct dataset preparation, preprocessing, and augmentation for both small and large

models.

• Perform hyperparameter tuning and model evaluation using appropriate metrics.

• Monitor and optimize model performance, inference speed, and memory footprint.

• Contribute to internal tools and libraries that streamline ML experimentation and deployment.

Requirements:

• Strong programming skills in Python and experience with ML frameworks like PyTorch,

HuggingFace Transformers, or scikit-learn.

• Experience in training and fine-tuning large language models using modern toolkits (e.g.,

PEFT, DeepSpeed, FSDP).

• Familiarity with distributed training, mixed-precision training, and checkpoint management.

• Solid understanding of ML fundamentals, including model selection, training/validation

workflows, and performance evaluation.

• Hands-on experience with data handling, feature engineering, and synthetic data generation.

• Working knowledge of experiment tracking tools (Weights & Biases).

• Familiarity with cloud environments (AWS, GCP, Azure) or on-premise GPU clusters.

• Strong problem-solving skills and ability to work independently in a fast-paced environment.

Nice to Have:

• Experience with retrieval-augmented generation (RAG), prompt tuning, or instruction tuning.

• Exposure to quantization and model compression techniques.• Experience with deploying ML models using APIs or serving frameworks (e.g., FastAPI, Triton

Inference Server).