We build on top of the best language models available to give every candidate the experience of a dedicated, thoughtful career agent - without the bottleneck of a single human.
Here's how we think about AI and where we use it.
Our principles for building with AI. Not aspirational - this is how we actually work.
We treat AI the way good engineers treat any tool - with clear expectations, measurable outcomes, and healthy skepticism. LLMs are powerful but imperfect. We design around their strengths and guard against their weaknesses.
Clera does not make hiring decisions. AI handles matching, research, and preparation at scale, but candidates choose where to apply and companies choose who to hire. Every AI recommendation comes with reasoning a human can review.
We use Claude, GPT, and Gemini depending on the task. We don’t marry a single provider - we pick the best model for each job and swap as the landscape evolves. Our value is in the system we build, not the model we call.
We aim to make our reasoning inspectable. Match explanations, scoring breakdowns, and preparation insights are surfaced to candidates wherever the system can produce them. We don’t hide behind "the algorithm decided."
No AI system is bias-free, including ours. We design for skills and trajectory rather than pedigree, monitor match distributions for patterns we didn’t intend, and treat fairness as continuous work - not a one-time fix.
We deploy new AI features quickly but measure everything. Tracing and evaluation are part of the pipeline - if a prompt change degrades match quality, we want to catch it fast. Speed and accountability aren’t at odds.
AI touches most of what Clera does - but always with clear purpose and measurable outcomes.
LLMs read and reason about your full profile - skills, experience, preferences, and career goals - to find opportunities that actually fit. Not keyword overlap. Real understanding of what you’re looking for and what a company needs.
For every matched role, AI generates personalized preparation: company context, likely interview topics, role-specific talking points, and honest feedback on where your profile is strong or could use more detail.
AI reads your resume the way a thoughtful hiring manager would - identifying strengths, gaps, and how your experience maps to specific roles. You get actionable feedback, not a generic score.
AI agents handle routine coordination, status updates, and follow-ups so nothing falls through the cracks. When a conversation needs nuance, humans step in. The system knows the difference.
We use AI to cut through vague job descriptions - extracting what you’d actually do day-to-day, what skills genuinely matter, and what the company culture looks like based on real signals, not marketing copy.
Every match outcome, every piece of feedback, every successful placement makes the system smarter. Not through retraining foundation models - through better prompts, better context retrieval, and better evaluation of what "good" looks like.
Practical choices, not hype.
Anthropic Claude, OpenAI GPT, Google Gemini - routed per task. We switch models when better options emerge.
Vercel AI SDK for tool-calling agents. Trigger.dev for background AI workflows. Custom prompt pipelines with structured outputs.
Ongoing tracing of LLM calls via Langfuse - latency, cost, quality. We review traces regularly so we can catch regressions before they reach users.
Typesense for fast candidate and job search. Structured retrieval over vector search - we prioritize precision over vibes.
Automated evals on match quality, prompt accuracy, and agent behavior. Manual review for edge cases. Both matter.
Zod-validated tool calls and JSON schemas everywhere a model returns data. We don’t parse free-form text - we constrain the model to shapes our code can trust.
Honest answers about how we use AI in recruiting.
Want to see AI-powered matching in action? It's free.