Appeals V2 - User Safety Experience

∗ Improved Snapchat’s review process by streamlining task creation and optimizing the review tool.

∗ Reduced turn-around-time for reviewing an appeal from 10 minutes to < 1 min, and scaled the platform to handle 14x task volume without loss in operational metrics.

[Tech Lead] Safety Enforcement Service - User Safety Experience

∗ Designed a new service to house the company’s core enforcement logic for moderating users and their content.

∗ Improved developer velocity by reducing the average testing turnaround time from 6 hours to 25 minutes.

My Reports - User Safety Experience

∗ Designed and implemented a feature enabling users to check the status of their submitted reports within the app.

∗ The project successfully achieved a key OKR aimed at improving user trust through visibility into our moderation process.

– CSAT survey - User Safety Experience

∗ Designed and implemented a process to assess the quality of Snapchatters’ experiences using our safety toolkit.

∗ Established baseline customer satisfaction metrics that serves as the team’s key success metric.

Abuse Detection:

• Developed a centralized platform for abuse detection and led the effort to build and onboard ML models for 4 different teams: Google Play reviews, Gmail, Android Messages and Child Safety.

• The platform supports models for both classification and ranking, and improved the quality of samples sent for manual review. This cut down operational expenditure and improved SLA.

• Coordinated efforts across multiple cross functional partners, including legal, privacy, researchers, engineers and PMs.

Sentiment Analysis:

• Built an ETL pipeline to run graph clustering on the scale of billion entities to categorize data for performing sentiment analysis on Google Cloud reviews.

• Reduced the time taken to analyze new incoming datasets from 6 months to 1 month, allowing us to address developer pain points quicker.

• Our efforts led to a patent application that is pending approval.

Semantic Parsing for Google Assistant:

• Improved language understanding for Google Assistant using semantic equivalence and nearest neighbor search.

• My effort led to a 0.08% increase in coverage which translates to 1000s of additional queries understood per day.

• Built an A/B testing infrastructure to evaluate parsing ability of various NLU models. My effort brought down the time taken to run experiments from 8 hours to 2 hours.

• Hosted an intern to improve the quality of our NLU models that parse queries related to alarms and timers. This effort increased coverage by an additional 0.03%.

Google iOS Search App Team:

• Led the effort to port context-based query functionality from Android to iOS Google Search app. This effort displays suggestions based on past search history and reduced time taken to formulate queries by 400ms/user.

• Led a cross functional effort to integrate existing A/B testing framework onto the iOS stack. My effort resulted in twice the number of A/B tests conducted in the year it was launched. The platform was also utilized to roll out risky features and caught two critical bugs that would have caused a service outage.