Machine Learning Data Engineer

kadence; Montreal (administrative region); On-site; CAD 120,000 - 180,000; Full time

Job summary

Kadence in Montreal is hiring multiple Senior Machine Learning Data Engineers to join an ambitious AI research team focused on training-data engineering and curation at web-scale. You’ll help transform raw web-scale data into high-quality datasets used to train next-generation models.

This role sits at the intersection of data engineering, ML, NLP, and large-scale training data curation, building scalable infrastructure, quality scoring, and tooling for researchers.

Qualifications

• 5+ years of experience across ML engineering, data processing, NLP, or data infrastructure.
• Experience with very large unstructured text datasets, ideally web/foundation-model scale.
• Strong production-level Python programming.
• Experience with distributed data-processing technologies such as Spark, Ray, Flink.
• Experience with PII scrubbing, content-safety filtering, dataset quality.

Responsibilities

• Design, build and scale pipelines that transform raw web-scale data into high-quality datasets for training large ML models.
• Process extremely large unstructured text corpora, including datasets at trillion-token scale.
• Build data-processing systems covering deduplication, model-based quality scoring, filtering, PII removal, metadata extraction.
• Develop automated data-quality systems using heuristics, ML classifiers, LLM evaluators and human-in-the-loop workflows.
• Build monitoring and guardrails to identify data-quality regressions before they impact training.
• Collaborate with AI researchers to understand evolving data requirements and identify gaps in datasets.
• Design robust contamination and data-leakage detection systems.
• Build internal tools to explore, query and understand large datasets.
• Optimize large-scale distributed processing for throughput, reliability and cost.

Skills

Python

Distributed data processing

Spark

Ray

Flink

Data pipelines

NLP

Machine learning engineering

PII scrubbing

Airflow

Tools

Spark

Ray

Flink

Airflow

Prefect

Dagster

Docker

Kubernetes

Terraform / IaC

Experiment tracking

Job description

This is an opportunity to work alongside a highly accomplished AI research team tackling fundamental problems in advanced machine learning.

Rather than maintaining established pipelines, you’ll be helping develop new approaches to training-data engineering and curation where established playbooks often do not yet exist.

The organization is building a substantial technical team in Montreal, and we are hiring multiple people across this area.

If you’ve worked on LLM training data, large-scale NLP pipelines, foundation-model infrastructure or web-scale data processing, I’d be interested in speaking with you.

We’re working with an ambitious AI research organization building next-generation machine learning systems and are looking for multiple Senior Machine Learning Data Engineers to join its growing technical team.

This role sits at the intersection of data engineering, machine learning, NLP, and large-scale training data curation.

You’ll be responsible for building and scaling the infrastructure that transforms raw, web-scale data into high-quality datasets used to train advanced AI models.

The core challenge is engineering data quality at enormous scale: developing filtering systems, model-based quality scoring, dataset transformations, contamination detection, and tooling that allows researchers to understand and work with training corpora effectively.

What You’ll Work On

• Design, build and scale pipelines that transform raw web-scale data into high-quality datasets for training large machine learning models.
• Process extremely large unstructured text corpora, including datasets at trillion-token scale.
• Build sophisticated data-processing systems covering: Deduplication, Model-based quality scoring, Heuristic filtering, Toxicity and content-safety filtering, PII detection and removal, Metadata extraction, Dataset transformations, Versioning and provenance tracking
• Develop automated data-quality systems using a combination of heuristics, ML classifiers, LLM-based evaluators and human-in-the-loop workflows.
• Build monitoring and guardrails to identify data-quality regressions before they impact downstream model training.
• Work closely with AI researchers to understand evolving training-data requirements and identify gaps within existing datasets.
• Design robust evaluation contamination and data-leakage detection systems.
• Build internal tools that allow researchers to easily explore, query and understand large datasets.
• Optimize large-scale distributed processing systems for throughput, reliability and infrastructure cost.

What We’re Looking For

• 5+ years of experience across machine learning engineering, data processing, NLP, data infrastructure or related areas.
• Experience working with very large unstructured text datasets, ideally at web or foundation-model scale.
• Strong production-level Python.
• Hands‑on experience with distributed data-processing technologies such as: Spark, Ray, Flink
• Experience designing and optimizing high-throughput distributed pipelines.
• Experience with areas such as: PII scrubbing, Content-safety filtering, Dataset quality, Evaluation contamination, Training-data governance
• Experience with workflow orchestration frameworks such as Airflow, Prefect or Dagster.
• Ability to work closely with both research and engineering teams and translate research requirements into scalable infrastructure.

Particularly Relevant Experience

We’d be especially interested in people who have worked on:

• Pretraining or post-training data pipelines for LLMs
• Common Crawl or comparable large-scale datasets
• LLM-based data-quality evaluation
• ML classifiers for filtering or scoring training data
• Dataset deduplication and contamination detection
• Distributed ML/data infrastructure

Experience with vLLM, SGLang, Docker, Kubernetes, infrastructure-as-code, experiment tracking or open-source NLP/data tooling would also be valuable.