AI Model Lead

Ideagen

9d•Remote

About The Position

Ideagen is building domain-specific AI models for environments where accuracy, traceability, and accountability matter. As AI Models Lead, you will take technical ownership of how language models are adapted, evaluated, and promoted into production across EHS, Quality, and GRC workflows. This is a senior, hands-on engineering leadership role with responsibility for setting the technical standard for fine-tuning at scale, from data and evaluation design through to production-ready adapter libraries. You will lead the AI Models Team and work closely with domain experts, platform engineers, and AI leadership to ensure models behave correctly in regulated industry contexts. This role combines deep technical execution with people leadership and offers a clear growth path as Ideagen’s AI organisation continues to scale.

Requirements

A proven track record of fine-tuning language models that serve real production traffic, with a strong understanding of operational failure modes
Deep practical experience with LoRA and QLoRA, and clear intuition for the trade-offs between quality, compute cost, and inference performance
Hands-on experience applying alignment techniques such as supervised fine-tuning and preference optimisation to real-world problems
Strong tooling fluency with modern ML frameworks and experimentation platforms, including debugging and scaling training pipelines
High-quality Python engineering skills, with emphasis on readable, testable, and maintainable training and evaluation code
Experience leading or mentoring engineers in a technical ML environment, with a people-first approach to leadership
Comfortable working with subject-matter experts to define “what good looks like” when evaluation requires real domain judgment

Nice To Haves

Background or interest in regulated domains such as EHS, quality management, healthcare, or financial services is advantageous

Responsibilities

Leading and developing the AI Models Team, setting engineering standards, mentoring engineers, and establishing a strong technical culture
Designing and owning the end-to-end fine-tuning methodology, including training configuration, evaluation gates, reproducibility standards, and promotion criteria
Making base model selection decisions across foundation model families, balancing capability, cost, and inference constraints for different domain tasks
Owning LoRA and QLoRA adaptation strategy, including configuration choices, training compute decisions, and compatibility with the inference layer
Applying supervised fine-tuning and preference optimisation techniques to domain-specific classification and compliance problems
Building robust evaluation frameworks with domain experts to assess whether models are performing correctly in real EHS, Quality, and GRC use cases
Leading dataset design and construction, defining labelling standards and ensuring training data is versioned, auditable, and reproducible
Owning production training and adapter lifecycle management, from large-scale training jobs through to registry promotion and release readiness