À propos de ce poste Senior Data Scientist chez Citizen
A Note From the Founder
Citizen makes thousands of decisions a day about what people see, when they are alerted, and what we build next. Every one of those decisions leaves a trail in our data: what people opened, what they followed, what kept them coming back. Few companies get to learn from a dataset like this, about moments that matter this much. I want a data scientist who loves that problem: someone who designs great experiments, finds the signal in real-world data, and helps decide what we build next, with AI tools that let one person cover more ground than a whole analytics team could a few years ago. If you want your work to shape what a city sees in the moments that matter most, come build it with us.
The Role
Citizen is the #1 safety app in the U.S. Every day, thousands of videos are captured live at the scene of real incidents and distributed in real time to people in every major American city. No other consumer product has this. It is a live network, a newsroom, and a safety tool at once, and it is growing again.
Underneath it is one of the most unusual datasets in consumer technology: a decade of incidents, live user video with location, every alert we have sent, and how millions of people respond to it, all instrumented as analytics events in BigQuery. As our senior data scientist, you will lead what we learn from that data: the experiments that decide what ships, the analysis that answers product questions, and the recommendation for what to try next.
You will work with a small, fast team across product, engineering, and operations. You will partner closely with the engineers shipping changes, Mission Control, and the Founder.
Why Now
Citizen is moving faster than at any point in its history. We are building AI into how incidents are written, who gets alerted and at what radius, and how people find and follow what is happening near them. Every one of those changes is a bet worth measuring well, and this role makes sure we learn from each one. It is also a great moment to be a data scientist: AI tools now handle much of the querying, segmenting, and first-draft analysis, so your time goes to the parts that need real judgment.
How We Work
You are a data scientist first: statistical judgment, rigor about what a number means, and healthy skepticism of a result that looks too good. Alongside that, you use AI tools and agents to write queries, build pipelines, and draft analyses, so you spend your time on the questions and the conclusions rather than the plumbing.
Engineering runs on a driver and builder model. Every initiative has one accountable driver who owns scope, sequencing, and tradeoffs, and a small set of builders who ship it with them. Drivers are chosen by who wants the ownership, not by seniority. You will contribute to other people's initiatives and lead your own early on.
What You Will Own
Experimentation. Our experimentation framework and the practice around it: hypothesis, assignment, power, sample size, and guardrail metrics; sample-ratio checks, variance reduction, novelty and seasonality effects, and segment effects; a clear readout format and cadence. Product, notification, and content changes that reach millions of people run through it.
Product analytics. The event model end to end, from Segment to BigQuery. Funnels, retention, engagement, and behavior by city and incident type. Answering the questions the team asks, and surfacing the ones it should be asking.
Turning results into decisions. Every readout ends with a recommendation: what the result means for users, what to build or test next, and what to stop.
Notifications and relevance. Who gets alerted, at what radius, for what. Notifications are the product's most powerful lever. You will own how they are measured and the experiments that improve them.
Core metrics. Partnering with the data team on the definitions behind the metrics leadership and the board use, and on the monitoring that keeps them reliable.
Measuring AI features. Partnering with ML and engineering on how we evaluate AI in the product from the user's side: what good looks like, and whether a change made the experience better.
Product visualization. Finding the few metrics that matter among many, and building clean, readable views of product health and experiment results that anyone in the company understands at a glance.
Analysis tooling. Extending the AI tools and agents we use to query, segment, and draft analysis, so the answer to a product question takes hours, not weeks.
Your First 30 Days
Days 1–10. Learn the product (use it every day in New York), the event model, and our metrics. Rebuild one core product metric from raw events and review a recent A/B test and its readout.
Days 11–20. Ship your first analysis that informs a product decision. Start your first initiative: strengthen the experimentation framework (design template, power checks, guardrails, readout format) and launch one test through it.
Days 21–30. Deliver your first readout with a recommendation, and present an experimentation plan for the next two quarters: the questions that matter most, the tests that answer them, and the metrics you plan to move.
How Success Is Measured
Experiments run per quarter, and the share whose readout leads to a decision
Results the team trusts: pre-registered metrics, clean assignment, and wins that hold up over time
Time from product question to answer, shrinking as AI tooling takes on more of the work
Notification relevance: opens, dismissals, and opt-outs per alert improving together
Clear metric definitions and event coverage across the core product
What You Bring
Required
5+ years in data science, product analytics, or experimentation, with at least 3 years on a consumer product at scale (millions of users)
A degree in statistics, economics, mathematics, computer science, or another quantitative field, or equivalent practical experience. An MS or PhD is welcome, not required.
Hands-on experience designing, running, and analyzing A/B tests: power and sample size, guardrail metrics, sample-ratio checks, and the common pitfalls (peeking, novelty effects, interference)
Strong applied statistics: regression, hypothesis testing, variance reduction (e.g. CUPED), and causal inference methods for when a clean test is not possible (difference-in-differences, matching, synthetic control)
Expert SQL and strong Python (pandas, NumPy, statsmodels or SciPy) in a modern cloud warehouse; we use BigQuery and Segment
A strong eye for data visualization: you can take a crowded set of metrics and turn it into a clean, readable chart that anyone on the team understands
A track record of analysis that changed what a product team built, and the communication skills to make that happen: clear written readouts and confident presentation to engineers and executives
Day-to-day use of AI tools in your own analytical workflow (e.g. LLM-assisted SQL and Python, agents for recurring analysis)
Preferred
Experience with notifications, ranking, recommendations, or other relevance problems
Experience with geospatial or location-based data
Experience with an experimentation platform (e.g. GrowthBook, Statsig, Optimizely, Eppo) or building one
Familiarity with event instrumentation and data modeling (dbt, event taxonomies, data quality monitoring)
Experience evaluating ML or LLM-powered product features from a user-outcome perspective
Consumer, media, or real-time product experience
Who You Are
Curious. Handed a question, you often come back with a better one.
Rigorous and practical: you know when a result is solid and when it needs another look, and you can explain why.
An owner who likes leading initiatives as well as contributing to them.
Motivated by the mission. Safety matters to real people in real moments, and you want your work to contribute.
Excited to work in person with the team in our New York office.
Compensation
$200,000 - $240,000 base, plus equity