$35–$45 CAD/hour · 10–15 hours/week · Remote within Canada

Toronto/GTA preferred; strong candidates across Canada welcome.

Solveion is a small independent AI consultancy based in Vaughan, Ontario. We’re hiring a part-time AI Model Trainer to evaluate large language models and help build reliable evaluation methods for real client projects.

You’ll assess model responses, investigate errors, and develop rubrics that make quality measurable. We’re particularly interested in people who can explain both why an answer fails and when the grading criteria need improvement.

What you’ll do

- Evaluate and rank responses for factual accuracy, reasoning, relevance, and instruction following.
- Create and refine evaluation rubrics, grading guidelines, and reference answers.
- Identify hallucinations, unsupported claims, logical errors, and recurring failure patterns.
- Write prompts and edge cases, research claims, and fact-check model outputs.
- Review evaluation work, resolve grading disagreements, and improve dataset quality.

What you’ll bring

- Solid analytical judgment, attention to detail, and clear written English.
- Experience in AI evaluation, research, fact-checking, or structured quality assurance.
- Ability to apply detailed guidelines consistently and explain ambiguous decisions.
- Availability for approximately 10–15 hours per week.
- Canadian work authorization and the ability to perform the work from Canada.





Experience we value Prior AI training, LLM evaluation, response ranking, human feedback/RLHF, or advanced annotation work through Handshake, Alignerr, Outlier, Mercor, DataAnnotation, TELUS Digital AI Community, or similar programs is especially relevant.

Rubric development, evaluator calibration, and reviewing other evaluators’ work are strong advantages.

Familiarity with Python, SQL, RAG, APIs, JSON, prompt engineering, or evaluation frameworks is helpful, not required. We value demonstrated judgment over a particular degree or platform.

Working with Solveion

You’ll work closely with a small team on client problems, with room to shape our evaluation methods and take on greater responsibility.

Some engagements have Canadian data-residency and data-sovereignty requirements. Work must take place in Canada using approved systems. You’ll follow client-specific confidentiality rules and keep client data out of personal AI accounts and unapproved storage.

How to apply

Apply through LinkedIn Easy Apply with your resume. Highlight your model-evaluation, rubric-development, and AI-training experience. Don’t include confidential work from previous clients or platforms.

Shortlisted candidates may complete a short evaluation exercise using synthetic examples, followed by an interview and a paid trial task.

📌 AI Model Trainer - Part-Time (Canada)
🏢 Solveion
📍 Canada