Skip to main content

shipped field record

Long-Term Impact Research

2025-2026Data Science & ResearchClient: TalkingPoints

I worked at TalkingPoints from 2022–2026. Each project lists its own dates, so some begin later within that span.

Evidence-based product development, impact measurement
Problem
Product decisions needed stronger evidence about long-term student outcomes.
My role
Built privacy-conscious analysis infrastructure and research datasets.
Shipped
Repeatable impact-analysis pipelines and conversation research tooling.
Proof · 2026-07
Moved impact conversations from intuition toward inspectable evidence.

Public systems

Research infrastructure for impact analysis. Clio-inspired conversation analysis pipeline. Dialogue segmentation, utterance role labeling, privacy-preserving NLP at scale.

Materials & technology

SQLPythonStatistical analysisSnowflake

Links

Frequently Asked Questions

How do you handle the challenge that correlation isn't causation?

Carefully. We're explicit about what the analysis can and can't prove. Correlation with controls for observable factors is a start, but it's not a randomized trial. The goal is to move from 'we think we help' to 'here's evidence that's consistent with helping.'

How does this research influence product decisions?

It shifts conversations from opinion to evidence. Instead of 'we think feature X helps,' we can say 'students whose families received feature X had Y% fewer absences, controlling for Z.' That's a different kind of argument.