Remote ML Engineer (NLP/LLM) for Risk Intelligence

techholding; United States; Remote; USD 120,000 - 180,000; Full time

Job summary

Techholding is seeking a senior ML engineer to develop NLP and LLM capabilities for a risk intelligence platform, turning large volumes of online text into actionable signals.

Based remotely in Mexico, you will work as an individual contributor delivering production ML models, with regular collaboration with U.S.-based engineering and product stakeholders.

Responsibilities include deploying scalable services, monitoring model performance, and improving data pipelines for multilingual sources.

Qualifications

• 4+ years building and deploying production ML models.
• Strong Python and ML framework experience.
• Experience with NLP tasks in real-world text.
• Experience with LLMs and retrieval-augmented generation.
• Experience deploying ML services on cloud with Docker and CI/CD.

Responsibilities

• Design, train, and deploy models for text classification and extraction.
• Develop LLM features via prompt design and RAG workflows.
• Build embedding pipelines and semantic search to surface threats.
• Deploy scalable ML services with monitoring and cost goals.
• Create evaluation data to track bias, errors, and drift.
• Improve ingestion and labeling pipelines for multilingual text.
• Monitor production models and retrain as needed.
• Join daily standups with the U.S.-based team.

Skills

Python

NLP

LLMs

Prompt engineering

Vector search

Model deployment

CI/CD

Tools

PyTorch

TensorFlow

Hugging Face Transformers

Docker

FAISS

Job description

Techholding is seeking a senior ML engineer to develop NLP and LLM capabilities for a risk intelligence platform, turning large volumes of online text into actionable signals.

Based remotely in Mexico, you will work as an individual contributor delivering production ML models, with regular collaboration with U.S.-based engineering and product stakeholders.

Responsibilities include deploying scalable services, monitoring model performance, and improving data pipelines for multilingual sources.