Charlie Proctor

Opioid overdose is the leading cause of death for Americans under 50. Resilient Lifescience is developing a solution: a wearable medical device that can automatically detect and reverse an opioid overdose. It’s a patch worn on the abdomen that monitors a user’s vitals and injects naloxone when it detects an overdose, removing the need for a bystander to administer the lifesaving drug.

During my time at Infinitus, we were focused on automating phone calls for health insurance benefit verification (BV). As one of the first engineers on the backend team, I collaborated directly with the founders, product leads, and our operations team across a variety of projects as we grew from 10 to 20 engineers and series A to B: - Scaling our offering to multiple customers required a way to efficiently validate and ingest BV tasks asynchronously. Expanding our customer-facing API, I built rule-based tools for resolving the ambiguity inherent in BV information collected from the provider. - As the volume of outbound calls increased, we needed a way to schedule them throughout the day. We had a fixed pool of resources to handle calls, and each inbound BV task came from the provider with a different deadline. The calls to complete these tasks involved insurance companies with variable open/close hours and hold times. Using a combination of heuristics and statistical modeling, I designed and implemented a system to automatically schedule and place these calls. - Running hundreds of calls simultaneously across distributed infrastructure required constant improvements to our system’s latency and fault tolerance, such as exponential backoff or cluster-wide caching. Logging, monitoring, and alerting proved crucial during system-wide outages. I had the opportunity to mentor an intern for 10 weeks during the summer of 2021. She worked on machine learning tools for the estimation of variables required by the scheduler - most notably, hold times when calling the insurance companies.

The click-to-call mobile ad format allows users to call an advertiser directly. Routed through Google, the call is tracked in order to provide attribution for the advertisers. Within this team, I worked on detecting and reporting various types of conversions and other call-related metrics to the advertisers. My projects included: - Adding support for a call-based charging model (as opposed to standard click-based models), - Updating data processing pipelines to display extension-level call reporting in AdWords, - Using TensorFlow-based machine learning models to identify high-fidelity conversions from call recording transcripts, and - Building an infrastructure tool for dynamically loading JavaScript workflows for call routing. While employed at Google, I conducted 120+ software engineering interviews using multiple questions across a variety of technical domains and programming languages. After each interview, I provided detailed written feedback and a recommendation to the hiring committee.