Senior Applied ML Scientist - 1086

Lightcast•Moscow, ID

31d

About The Position

Lightcast is seeking a Senior NLP / Machine Learning Scientist to join its AI/ML Lab and build state-of-the-art NLP models for large-scale labor market data. The role develops and productionizes models that transform unstructured data, such as job postings and professional profiles, into structured labor market intelligence. Core problem areas include classification, information extraction, entity recognition, normalization, semantic matching, and multilingual NLP. The position spans the full model lifecycle, from experimentation and fine-tuning through evaluation, optimization, and production inference at scale. The scientist partners with ML and data engineers to deploy scalable models and training pipelines in AWS, contributes to modeling strategy and technical standards, and mentors team members.

Requirements

Master's degree or PhD in Computer Science, Machine Learning, NLP, or a related field.
Strong experience building and deploying production NLP or ML systems.
Deep understanding of transformers, embeddings, classification, information extraction, and entity normalization.
Strong Python and PyTorch experience, with hands-on experience using Hugging Face or similar frameworks.
Experience with large-scale model training, evaluation, and productionization.
Strong problem-solving skills and ability to independently drive ambiguous ML problems from research to production.
Hands-on experience fine-tuning LLMs or foundation models.
Experience with LoRA, QLoRA, PEFT, instruction tuning, or model distillation.
Experience with model inference optimization, including quantization, ONNX, TensorRT, vLLM, TorchCompile, or GPU optimization.
Experience with multilingual NLP, semantic search, retrieval, entity linking, or knowledge graphs.
Experience with AWS, SageMaker, Docker, MLflow, or similar MLOps tooling.

Responsibilities

Build NLP models for classification, extraction, normalization, entity linking, and semantic matching.
Develop transformer-based and embedding-based models, including architectures such as BERT, DeBERTa, ModernBERT, Sentence Transformers, and modern foundation models.
Fine-tune and adapt LLMs using methods such as supervised fine-tuning, LoRA, QLoRA, PEFT, and distillation.
Evaluate whether to use LLMs or smaller task-specific models based on accuracy, latency, throughput, and cost.
Optimize models for production using quantization, distillation, batching, model compilation, and efficient inference frameworks.
Develop multilingual NLP models and cross-lingual approaches for global labor market data.
Design robust evaluation frameworks, perform error analysis, and continuously improve model quality.
Partner with ML and data engineers to deploy scalable models and training pipelines in AWS.
Stay current with advances in NLP, LLMs, embeddings, and efficient machine learning, and translate promising research into production systems.
Mentor team members and contribute to technical standards and modeling strategy.