Job DescriptionYou will join a search team responsible for a large-scale production search system combining lexical retrieval (Solr/Lucene) with semantic and ML-based ranking. The environment is mature and heavily instrumented, with ranking models that have been in production and continuously refined for years.
Your role is to bring rigorous statistical and causal reasoning to complex questions around ranking performance, experimentation and user behavior — separating causation from correlation and quantifying uncertainty in production decisions.
This is fundamentally a statistics and applied mathematics role. You will work closely with Search and ML Engineers, with a focus on experimental design, causal inference, variance reduction and statistical rigor rather than production software engineering.
Responsibilities
- * Design experiments (A/B tests, interleaving, quasi-experiments) for ranking and relevance changes, including power analysis, randomization strategy, and metric selection.
- Apply causal inference methods (e.g., difference-in-differences, instrumental variables, propensity/matching methods, synthetic control, uplift modeling) to observational search and engagement data where randomized experiments are impractical.
- Diagnose confounding, selection bias, and interference (e.g., network/spillover effects between ranked results) in search experiments and metrics.
- Partner with ML and search engineers to interpret model behavior statistically — separating genuine signal from noise, seasonality, novelty effects, and measurement artifacts.
- Develop and validate statistical models of user behavior and search outcomes (e.g., click models, exposure models, position bias correction) to support offline evaluation.
- Communicate findings and their statistical caveats clearly to engineering and product stakeholders, including what conclusions are and are not supported by the data.
- Advise on experimentation infrastructure and methodology (metric definitions, guardrails, sample ratio mismatch checks) to keep the team's experimental culture rigorous over time.
Must-have Qualifications
- * Advanced degree (MS/PhD or equivalent experience) in Statistics, Biostatistics, Econometrics, Applied Mathematics, or a related quantitative field.
- Demonstrated depth in causal inference: experimental design, quasi-experimental methods, and the assumptions/limitations underlying each.
- Strong grounding in classical statistical modeling — hypothesis testing, regression, variance estimation, Bayesian methods, or related techniques — with the ability to justify model choices mathematically.
- Experience analyzing behavioral or observational data at scale (e.g., search logs, clickstream, recommendation/engagement data) and reasoning about biases specific to that data (position bias, selection bias, feedback loops).
- Comfort working with real, messy production data and communicating uncertainty honestly rather than overclaiming significance.
- Sufficient programming/scripting ability (e.g., Python, R) to conduct independent analysis — engineering-grade software skills are not required.
- Excellent written and verbal communication, including the ability to explain statistical reasoning and caveats to non-statistician engineering audiences.
Extra Merit Qualifications
- * Prior experience in search, recommendation, or online marketplace settings where ranking decisions are evaluated via experimentation.
- Familiarity with information retrieval evaluation concepts (relevance judgments, ranking metrics such as NDCG/MRR) sufficient to collaborate with ML engineers, even without building such systems yourself.
- Experience with click/exposure modeling or position-bias correction methods specific to ranked
lists.
- Publication or applied track record in causal inference methodology (e.g., work with DoWhy, EconML, CausalImpact, or comparable frameworks/literature).
- Background collaborating with engineering teams without owning production code delivery.
Information
- Language Requirements: Fluent English (mandatory), Swedish
- Scope: 100%
- Location: Malmö, 3 days on site
- Start: Immediate
- Duration: 12 months, with possible extension
