Machine Learning Consultant

The role

We're building a production-grade Data & ML platform for a major Canadian telecommunications and digital-identity client, the system that will power real-time trust and fraud scoring across the market. It's a federated, privacy-first platform on AWS: data that be sourced into the platform or data that can't leave its source, a medallion Lakehouse, streaming and batch ingestion, and an ML-Ops spine that keeps models honest in production.

We're looking for a hands-on platform architect to be the senior technical anchor for this work, onshore, deeply technical, and comfortable working directly with the client's Chief Data & AI Officer.

What the platform is really for

At its core, this platform industrializes change across two lifecycles:

• Data pipelines: stand up new pipelines, and absorb changes to existing ones, with speed, low human touch (automation, templating, CI/CD), and quality built in (automated testing, data-quality gates, lineage).
• Models: deploy new models, monitor the ones in production, and absorb changes to existing ones, just as fast and safely (reproducible pipelines, automated evaluation and promotion, drift detection and retraining).

What you'll do

• Own the platform architecture end to end, ingestion, the medallion/lakehouse, feature and model pipelines, serving, and the governance and observability that hold it together, and make the trade-offs (which AWS services, what to deliberately not build) with clear cost and operability reasoning.
• Build the ML-Ops spine as a change-engine — reproducible training pipelines, a governed model registry, automated evaluation and controlled deployment, drift detection and retraining, so new models ship fast, production models are monitored, and model changes are absorbed with low human touch and high confidence.
• Engineer a data-pipeline factory — templated, automated pipelines (batch-first evolving to selective real-time, incremental and idempotent) so new pipelines stand up quickly and changes to existing ones flow through with quality gates and lineage intact, across an environment where data residency and federation are first-class constraints.
• Be the client's technical peer — walk their senior data/AI leaders through the architecture, engage on the hard questions (federated learning, edge inference, latency budgets, privacy-preserving signal exchange), and earn trust through substance.
• Set the team up to own it — this is a Build-Operate-Transfer engagement, so you'll build for handover to a lean in-house team, not for permanent dependency.

You build for the people who'll live in the platform

The best platform architects have walked in their users' shoes. You understand a data engineer's day, the 2 a.m. pipeline failure, the schema that drifted, the incremental load that double-counted, and you build tooling and guardrails that make those problems rare and recoverable. You understand a data scientist's day, needing to reproduce an experiment from three weeks ago, wanting a feature store that just works, watching a model quietly drift, and you build the platform that lets them move fast without breaking governance.

What we're looking for (must-have)

• Strong, production AWS experience across the data/ML stack — S3, Glue, SageMaker, streaming (Kinesis / MSK), and the surrounding services, with the judgment to choose the smallest set that meets the requirement.
• Real, in-production MLOps — training pipelines, model registry, deployment, monitoring, drift/retraining, not just model development.
• Data pipeline engineering at scale — batch and streaming, Lakehouse/medallion patterns, data quality, lineage.
• Architecture-level judgment — you can design a system, cost it, defend it, and adapt it.
• Onshore in Canada, and genuinely strong at client-facing technical communication, able to be the senior technical voice with a CTO or Chief Data & AI Officer.

Bonus points (nice-to-have)

• Federated / privacy-preserving ML (federated learning, secure aggregation, edge inference), or privacy-first signal-exchange architectures.
• Telecom, digital identity, fraud, or risk domain experience.
• Iceberg / open lakehouse table formats; infrastructure-as-code; multi-cloud (GCP/on-prem) exposure.
• Hands-on empathy from having been a data engineer or data scientist earlier in your career.