ClickFrame
In progressAn attribution and measurement pipeline for creators, built as a full data product end to end.
The interesting part is the measurement: tying spend to outcomes cleanly enough that a creator can trust the number and act on it.
I build and own data functions end to end, and I can tell you why the number moved.
Full-stack data scientist with eight years across the stack, from ETL pipelines to causal inference to production ML. My applied-economics background means I explain why, not just what, then build the system that acts on it. Open to full-time product and experimentation roles.
Selected experience
Eight years owning whole data functions on small teams, from pipelines to experiments to production models. A few results, outcome first.
- Product Hunt. Ran [PLACEHOLDER: N] experiments across the product. One [PLACEHOLDER: what it informed] and moved [PLACEHOLDER: which metric] by [PLACEHOLDER: figure]. I cared about which lever moved the metric, not just the lift.
- The Daily Wire. Rebuilt the analytics stack on dbt and Snowflake and shipped an LTV model that [PLACEHOLDER: the decision it drove or the result], turning a pile of disconnected sources into numbers the team could act on.
- Medimap. [PLACEHOLDER: result first] and cut [PLACEHOLDER: what] cost by [PLACEHOLDER: figure] by [PLACEHOLDER: the mechanism].
- WEC. Built ML [PLACEHOLDER: which systems] that [PLACEHOLDER: the outcome], early end-to-end ownership from data to deployed model.
Featured projects
Real, deployed things, shown as they develop. Full writeups on each project page.
Library RAG Search
LiveA retrieval-augmented search tool over a real library collection, built for an actual stakeholder.
Hybrid retrieval is the lever: combining dense and sparse search so a vague question still lands on the right passage.
Causal YouTube Thumbnail Study
In progressA causal study of how thumbnail choices affect click-through, done properly rather than by correlation.
The whole point is causal identification: separating what the thumbnail did from what the video, channel, and timing did.
About
I am a full-stack data scientist. For eight years I have owned whole data functions on small teams at WEC, Product Hunt, Medimap, and The Daily Wire. The stack changed by company, dbt pipelines, Snowflake, A/B testing, network analysis, production ML, internal Streamlit tools, but the job was the same: turn a mess of inputs into a clear picture leaders can act on, then build the system that keeps it honest.
My background is in applied economics, so causal thinking and problem framing come first. The question matters more than the model. Most teams asking for a better dashboard do not have a dashboard problem. They have a trust problem, and the fix is knowing why a number moved, not just that it did.
Right now I am building in public. ClickFrame and a handful of smaller projects are where I work out ideas in the open, from an attribution pipeline to a causal study of what actually drives clicks. I put them here as evidence of range and how I think, not as ventures I am pitching. I was also vetted into Toptal's top three percent, an outside read on the same skills.
If you are hiring for product or experimentation data science, I would like to talk.
Book a 30-minute intro call