The Role
Clearspace, the Y Combinator W23 company building interventions for compulsive phone use, is hiring an ML-focused engineer to own production models for classifying network traffic. The job ($150K-$200K salary, 0.5-1% equity, San Francisco-based, US visa holders only) requires experience with sequential models and time-series data.
The company wants someone who thinks beyond model architecture—about data volume, featurization strategy, and inference requirements. Specific experience with network traffic is a plus.
What This Signals
Clearspace currently uses Apple's ScreenTime API to trigger interventions (breathing exercises, usage budgets, teammate notifications). This hire suggests a pivot toward processing and filtering network traffic based on natural language rules—moving from surface-level app blocks to traffic-level behavioral classification.
The technical challenge: building models that classify compulsive behavior patterns in real-time network data. That's a step beyond simple screen-time tracking into predictive behavioral modeling.
The Context
The company claims 25 million hours saved across users, with 86% daily success rates and 97% satisfaction. Users report halving screen time permanently. These are founder-reported metrics without independent verification.
Clearspace's premise—that technology to protect attention should match the sophistication of platforms exploiting it—is reasonable. Social platforms deploy recommendation systems and attention engineering at scale. Consumer-side tooling has been mostly analog (willpower) or simplistic (timer apps).
The question is implementation. Network traffic classification for behavioral inference is technically ambitious, especially on iOS where Apple tightly controls network-level access. The job posting doesn't clarify whether this runs on-device or requires server-side processing—a distinction that matters for both privacy and performance.
What to Watch
Three things:
- Whether they can ship traffic-level classification on iOS without violating Apple's terms
- If the behavioral models prove more effective than existing ScreenTime API interventions
- How they handle the privacy implications of processing user network traffic
Digital wellness is a real enterprise concern as device addiction impacts productivity. But technical complexity and privacy trade-offs will determine whether this approach scales beyond early adopters. The hire is a signal of ambition. Delivery will be the test.