Production SLM Trainer & Model Evaluation Specialist

Jobtailor; Philippines; On-site; PHP 900,000 - 1,500,000; Full time

Job summary

Jobtailor is seeking a mid-level ML engineer to train and fine-tune Small Language Models for production systems. You’ll own evaluation, monitoring, and governance tasks, working with Claude Code, Postgres, and Rust, and documenting results for non-technical stakeholders.

You’ll collaborate with a junior developer, manage your own workstream in Jira, contribute to the GitHub codebase, and proactively flag risks related to model behavior, data handling, and security.

Qualifications

• Hands-on experience training and fine-tuning SLMs for production systems.
• Strong understanding of evaluation, monitoring, and regression testing.
• Familiarity with agentic AI systems and SLM usage.
• Data governance, access controls, and versioning knowledge.
• Proficient in Python and ML tooling.

Responsibilities

• Train, fine-tune, and improve SLMs under senior lead.
• Build evaluation and regression-test suites to catch drift.
• Monitor live model behavior in production and triage incidents.
• Curate and manage training/evaluation datasets per governance.
• Document model versions, decisions, results, and limitations.
• Collaborate with junior developer on the pod to keep docs/test coverage.
• Manage your own workstream via Jira and contribute to GitHub codebase.
• Flag risks around model behavior, bias, or data handling promptly.

Skills

Small Language Model Training

Model Evaluation & Monitoring

Data Handling & Governance

Python Programming

Independent Judgment

Tools

Claude Code

Postgres

Rust

Jira

GitHub

Job description

Jobtailor is seeking a mid-level ML engineer to train and fine-tune Small Language Models for production systems. You’ll own evaluation, monitoring, and governance tasks, working with Claude Code, Postgres, and Rust, and documenting results for non-technical stakeholders.

You’ll collaborate with a junior developer, manage your own workstream in Jira, contribute to the GitHub codebase, and proactively flag risks related to model behavior, data handling, and security.