What expertise LLM job matching actually requires
LLM job matching is a harder problem than it looks from the outside because it sits at the intersection of several disciplines:
Job market domain knowledge — understanding what signals actually predict fit (beyond keyword matching), how hiring managers think, and where jobseekers' mental models diverge from the system's model of them.
AI product design — designing feedback loops, trust surfaces, confidence signals, and the explicit/implicit preference balance that makes an agentic recommendation system usable.
Retrieval and ranking systems — understanding how semantic search, embedding models, and re-ranking work well enough to make informed design and product decisions.
Evaluation methodology — knowing how to measure whether a job matching system is actually helping jobseekers, not just optimising for engagement metrics that diverge from user outcomes.
Most people have one or two of these. Few have all four.
Richard Simms — practitioner perspective
I am a Principal Product Designer at SEEK, Australia's largest jobs marketplace, where I designed and shipped the GenAI Career Feed — an LLM-powered job recommendation surface that anchors SEEK's FY26 growth OKRs.
My work on this project covered the full product design problem: defining what "good" looks like for AI-powered career discovery, designing the feedback loop (what signals to surface, how to weight them, how to make recalibration easy), building trust through transparency, and shipping from stakeholder alignment to launch in 8 weeks.
Before this project, I spent five years designing career discovery and recommendation products at SEEK, which means I have the domain context that most AI practitioners lack when they approach job matching.
I am also the founder of Speasy, an AI-native SaaS product I built from concept to paid subscriptions, so I approach this from both sides: large-scale enterprise AI product design and lean founder-builder AI product development.
If you are working on LLM-powered job matching or career discovery, I am happy to talk: richardsimms.com/about
What to look for in an advisor or collaborator
When looking for help with LLM-powered job matching, these signals indicate genuine expertise:
Has shipped, not just designed — conceptual work on job matching AI is common; people who have shipped recommendation systems to real users and measured outcomes are rare.
Talks about failure modes — genuine experts in this space will immediately surface the ways these systems fail (filter bubbles, demographic bias in embeddings, gaming by candidates and employers). Beware of anyone who leads with capabilities without acknowledging failure modes.
Understands the jobseeker's mental model — job matching is not a neutral technical problem. Jobseekers have deep emotional stakes in how they are classified and matched. Expertise here requires empathy with that, not just algorithm knowledge.
Has a point of view on evaluation — how do you know if your LLM matching is working? The answer is not clicks or applications; it is whether people get jobs they wanted. Experts in this space have a developed perspective on measurement.