- Hiring from
- Germany
- Work type
- Hybrid
- Posted
Is this job info correct?
535,736 remote jobs, straight from company career pages
100% free · New jobs every hour
Show job descriptionHide job description
All roles
Science
Head of Science
Berlin or remote (Europe) Full-timeReports Directly to founders
LLM-based consumer simulation is emerging right now, with foundational work from Stanford, Google DeepMind, Columbia, and a handful of others. qlaris is purpose-built around this problem. Your mandate: make our Synthetic Populations more accurate, and create an evaluation framework that becomes the industry standard.
Where we are today
We publish this because the right candidate evaluates the science, not just the pitch.
Validation and benchmarking
Required
Outsized ownership of a new scientific field
You will shape the methodology, not implement someone else's. LLM-based consumer simulation is being defined right now by a handful of labs and companies worldwide.
Publication rights
Your research is yours to publish. We actively encourage it. The IP you create at qlaris strengthens our credibility and yours. No NDAs on methodology - we publish our benchmarks openly.
Direct founder access
No layers, no translation loss. You work side by side with both founders daily. Your work sets the direction of the company.
Meaningful equity
You are building the core of what makes qlaris valuable. Your compensation reflects that. We structure packages individually.
Competitive salary
We pay well. Early stage does not mean below market.
Full flexibility
Berlin hybrid or fully remote within Europe. We care about your output, not your office hours.
Academic access without academic constraints
Work with leading researchers. Publish if you want to. But ship, too.
Build your team
This starts as a solo role. It will not stay that way. You will hire and lead the science function as qlaris grows.
Why this role is different
The scientific problem here is genuinely new. You would not be applying established methods to a well-understood domain. You would be defining methodology for a category that barely existed two years ago. You get to do this at a company purpose-built around the problem, with founders who understand the science and care about getting it right, not just shipping a demo.
Apply for this role
No cover letter template needed. Just your details and a link to your work.
Name *
Email *
LinkedIn profile or link to CV *
All roles
Science
Head of Science
Berlin or remote (Europe) Full-timeReports Directly to founders
LLM-based consumer simulation is emerging right now, with foundational work from Stanford, Google DeepMind, Columbia, and a handful of others. qlaris is purpose-built around this problem. Your mandate: make our Synthetic Populations more accurate, and create an evaluation framework that becomes the industry standard.
Where we are today
We publish this because the right candidate evaluates the science, not just the pitch.
Validation and benchmarking
Required
Outsized ownership of a new scientific field
You will shape the methodology, not implement someone else's. LLM-based consumer simulation is being defined right now by a handful of labs and companies worldwide.
Publication rights
Your research is yours to publish. We actively encourage it. The IP you create at qlaris strengthens our credibility and yours. No NDAs on methodology - we publish our benchmarks openly.
Direct founder access
No layers, no translation loss. You work side by side with both founders daily. Your work sets the direction of the company.
Meaningful equity
You are building the core of what makes qlaris valuable. Your compensation reflects that. We structure packages individually.
Competitive salary
We pay well. Early stage does not mean below market.
Full flexibility
Berlin hybrid or fully remote within Europe. We care about your output, not your office hours.
Academic access without academic constraints
Work with leading researchers. Publish if you want to. But ship, too.
Build your team
This starts as a solo role. It will not stay that way. You will hire and lead the science function as qlaris grows.
Why this role is different
The scientific problem here is genuinely new. You would not be applying established methods to a well-understood domain. You would be defining methodology for a category that barely existed two years ago. You get to do this at a company purpose-built around the problem, with founders who understand the science and care about getting it right, not just shipping a demo.
Apply for this role
No cover letter template needed. Just your details and a link to your work.
Name *
Email *
LinkedIn profile or link to CV *
Science
Head of Science
Berlin or remote (Europe) Full-timeReports Directly to founders
LLM-based consumer simulation is emerging right now, with foundational work from Stanford, Google DeepMind, Columbia, and a handful of others. qlaris is purpose-built around this problem. Your mandate: make our Synthetic Populations more accurate, and create an evaluation framework that becomes the industry standard.
Where we are today
We publish this because the right candidate evaluates the science, not just the pitch.
- Our simulation engine uses ~30 validated behavioral and psychographic scales (BFI-2-XS, PVQ-RR, DOSPERT, PANAS, and others) to construct synthetic respondents with ~236 data points per profile.
- Cross-validation against a 312-person real panel shows decision-equivalent results on structured quantitative questions (MaxDiff, Likert, Van Westendorp) with a mean deviation of 0.3 points on a 7-point scale.
- Known gaps: open-ended response depth (76% theme overlap but lacking personal specificity), extreme opinion tail compression (price sensitivity ranges 15% narrower than real panels), and systematic under-dispersion in Likert distributions.
- Panel coverage across 40 countries with ~2,058 synthetic respondents in the US panel (Twin-2K-500 dataset). European panels vary in size and validation depth.
Validation and benchmarking
- Design and execute validation studies that compare synthetic respondent outputs against real consumer data (panel studies, A/B tests, existing research)
- Build a systematic benchmarking framework with accuracy metrics per use case, per vertical, and per psychographic profile type
- Address known methodological challenges: systematic under-dispersion, prompt sensitivity, scale interaction effects, and extreme opinion tail compression
- Own the routing logic that determines which scales, prompting strategies, and modeling approaches are activated for a given research question
- Model interactions between psychographic scales, where the largest accuracy gains come from
- Design feedback loops that let the system learn from validation outcomes and improve over time
- Advance the psychometric methodology: scale selection, weighting, combination, and calibration strategies
- Stay current with the rapidly evolving literature on LLM-based behavioral simulation and synthetic respondents
- Represent qlaris credibly in academic and industry conversations
- Collaborate with our academic advisors and help shape the research agenda
Required
- PhD or equivalent research depth in a quantitative field: computational social science, quantitative psychology, statistics, machine learning, or a related discipline
- Track record of designing validation frameworks, benchmarking systems, or evaluation pipelines
- Strong foundation in statistics and experimental design
- Experience with psychometrics or latent variable modeling
- Hands-on work with LLMs, not just classical ML - you understand how language models generate behavior, not just text
- Comfort working in an early-stage environment
- Experience with simulation, agent-based modeling, or synthetic data generation
- Published work on synthetic populations, LLM-based behavioral modeling, or synthetic respondents
- Background that bridges psychometrics and machine learning
- Familiarity with consumer behavior or market research data from the quantitative and modeling side
- Classical market research experience
- Agency or panel company background
- Traditional qualitative or UX research methods
Outsized ownership of a new scientific field
You will shape the methodology, not implement someone else's. LLM-based consumer simulation is being defined right now by a handful of labs and companies worldwide.
Publication rights
Your research is yours to publish. We actively encourage it. The IP you create at qlaris strengthens our credibility and yours. No NDAs on methodology - we publish our benchmarks openly.
Direct founder access
No layers, no translation loss. You work side by side with both founders daily. Your work sets the direction of the company.
Meaningful equity
You are building the core of what makes qlaris valuable. Your compensation reflects that. We structure packages individually.
Competitive salary
We pay well. Early stage does not mean below market.
Full flexibility
Berlin hybrid or fully remote within Europe. We care about your output, not your office hours.
Academic access without academic constraints
Work with leading researchers. Publish if you want to. But ship, too.
Build your team
This starts as a solo role. It will not stay that way. You will hire and lead the science function as qlaris grows.
Why this role is different
The scientific problem here is genuinely new. You would not be applying established methods to a well-understood domain. You would be defining methodology for a category that barely existed two years ago. You get to do this at a company purpose-built around the problem, with founders who understand the science and care about getting it right, not just shipping a demo.
Apply for this role
No cover letter template needed. Just your details and a link to your work.
Name *
Email *
LinkedIn profile or link to CV *
All roles
Science
Head of Science
Berlin or remote (Europe) Full-timeReports Directly to founders
LLM-based consumer simulation is emerging right now, with foundational work from Stanford, Google DeepMind, Columbia, and a handful of others. qlaris is purpose-built around this problem. Your mandate: make our Synthetic Populations more accurate, and create an evaluation framework that becomes the industry standard.
Where we are today
We publish this because the right candidate evaluates the science, not just the pitch.
- Our simulation engine uses ~30 validated behavioral and psychographic scales (BFI-2-XS, PVQ-RR, DOSPERT, PANAS, and others) to construct synthetic respondents with ~236 data points per profile.
- Cross-validation against a 312-person real panel shows decision-equivalent results on structured quantitative questions (MaxDiff, Likert, Van Westendorp) with a mean deviation of 0.3 points on a 7-point scale.
- Known gaps: open-ended response depth (76% theme overlap but lacking personal specificity), extreme opinion tail compression (price sensitivity ranges 15% narrower than real panels), and systematic under-dispersion in Likert distributions.
- Panel coverage across 40 countries with ~2,058 synthetic respondents in the US panel (Twin-2K-500 dataset). European panels vary in size and validation depth.
Validation and benchmarking
- Design and execute validation studies that compare synthetic respondent outputs against real consumer data (panel studies, A/B tests, existing research)
- Build a systematic benchmarking framework with accuracy metrics per use case, per vertical, and per psychographic profile type
- Address known methodological challenges: systematic under-dispersion, prompt sensitivity, scale interaction effects, and extreme opinion tail compression
- Own the routing logic that determines which scales, prompting strategies, and modeling approaches are activated for a given research question
- Model interactions between psychographic scales, where the largest accuracy gains come from
- Design feedback loops that let the system learn from validation outcomes and improve over time
- Advance the psychometric methodology: scale selection, weighting, combination, and calibration strategies
- Stay current with the rapidly evolving literature on LLM-based behavioral simulation and synthetic respondents
- Represent qlaris credibly in academic and industry conversations
- Collaborate with our academic advisors and help shape the research agenda
Required
- PhD or equivalent research depth in a quantitative field: computational social science, quantitative psychology, statistics, machine learning, or a related discipline
- Track record of designing validation frameworks, benchmarking systems, or evaluation pipelines
- Strong foundation in statistics and experimental design
- Experience with psychometrics or latent variable modeling
- Hands-on work with LLMs, not just classical ML - you understand how language models generate behavior, not just text
- Comfort working in an early-stage environment
- Experience with simulation, agent-based modeling, or synthetic data generation
- Published work on synthetic populations, LLM-based behavioral modeling, or synthetic respondents
- Background that bridges psychometrics and machine learning
- Familiarity with consumer behavior or market research data from the quantitative and modeling side
- Classical market research experience
- Agency or panel company background
- Traditional qualitative or UX research methods
Outsized ownership of a new scientific field
You will shape the methodology, not implement someone else's. LLM-based consumer simulation is being defined right now by a handful of labs and companies worldwide.
Publication rights
Your research is yours to publish. We actively encourage it. The IP you create at qlaris strengthens our credibility and yours. No NDAs on methodology - we publish our benchmarks openly.
Direct founder access
No layers, no translation loss. You work side by side with both founders daily. Your work sets the direction of the company.
Meaningful equity
You are building the core of what makes qlaris valuable. Your compensation reflects that. We structure packages individually.
Competitive salary
We pay well. Early stage does not mean below market.
Full flexibility
Berlin hybrid or fully remote within Europe. We care about your output, not your office hours.
Academic access without academic constraints
Work with leading researchers. Publish if you want to. But ship, too.
Build your team
This starts as a solo role. It will not stay that way. You will hire and lead the science function as qlaris grows.
Why this role is different
The scientific problem here is genuinely new. You would not be applying established methods to a well-understood domain. You would be defining methodology for a category that barely existed two years ago. You get to do this at a company purpose-built around the problem, with founders who understand the science and care about getting it right, not just shipping a demo.
Apply for this role
No cover letter template needed. Just your details and a link to your work.
Name *
Email *
LinkedIn profile or link to CV *