Cameron Zabriskie

Experience

ZONTAL

• Designed the autoscaling for the product's ~15 interactive viewer applications, moving them off a fixed-size shared node group onto a dedicated Karpenter node pool. Chose a combined CPU-and-session scaling signal after load testing showed either alone missed real cases, and sized the pool with a degradation-ratio model validated at 2,500 seats.
• Found and cleared cloud spend nobody was tracking: 3 TB of build artifacts deleted with automatic cleanup rules so it stops recurring, 82% of the account's load-balancer target groups identified as orphaned, and a runaway scheduler burning $37.89/day.
• Set the company's first service level objectives for customer-facing software, covering how fast it loads, how responsive it stays, and how often errors are acceptable, then measured every service against them so leadership could see where the product stood and what it would take to close the gap.
• Ran ~25 production incident investigations, each written up as symptom, root cause, and fix: cluster and authentication failures, automation that stranded whole environments, customer database outages, capacity shortfalls, and CI slowdowns across the engineering org.
• Built and deployed the internal MCP server that lets AI coding agents work against the platform directly, a shared service the team registers with their editor, scoped to the host's IAM role and logging every call for an audit trail. Wrote the review gate that goes with it: tests and architecture review before anything ships.

ZONTAL

• Built the pipeline behind what the company sells: parsers that map raw instrument output (mass spectrometry, NMR, PCR, flow cytometry) into a vendor-neutral schema instead of a format per instrument, the ingest that loads it into a document database, and the web applications scientists use to explore their results.
• Moved the company's public website off an agency-hosted WordPress stack onto Cloudflare Pages and Workers we run ourselves, cutting hosting to about $47/month and ending a process where content changes waited up to three months on the agency. Marketing now drafts and stages content themselves, with engineering only security-reviewing the push to production.
• Built the in-house replacement for a licensed third-party tool that converted documents (Word, Excel, PowerPoint, etc.) to PDF, removing the license cost and a dependency we couldn't change on our own schedule.
• Partnered with sales on demos and proofs of concept: hosted a legacy vendor chromatography system on AWS so a prospect could watch their own software push data into our platform, and built web applications showing how their data would be handled, including parsing encoded instrument formats.
• Gathered requirements directly from scientists at nine pharma and biotech customers and turned them into specs the team could build against. Ran scrum for multiple developer teams and mentored junior engineers into owning customer-facing work.

Utah State University

• Designed and teach a semester-long cloud computing course covering IAM and security, Docker and ECR, EC2, Lambda with EventBridge and API Gateway, ECS/Fargate, EKS, CloudFormation and AWS CDK, RDS/Redshift/DynamoDB, SQS/SNS, and CloudWatch observability.
• Built the curriculum, lecture videos, and hands-on assignments from scratch, with interview-style assessments aimed at making the material transfer directly to students' jobs.
• Manage teaching assistants and run open office hours across the semester.

Biolexis Therapeutics

• Took the company's heaviest computational work, the small-molecule protein docking and molecular dynamics simulations behind its drug discovery, from weeks down to under a day by moving it onto distributed compute (Ray on Anyscale), turning experiments that had been impractical to run into routine ones.
• Built the ELT pipelines that pulled research data into graph, relational, and NoSQL stores, then put APIs in front of them (Lambda and API Gateway) so other teams and applications could pull what they needed instead of asking someone to extract it.
• Replaced manually assembled AWS environments with infrastructure as code (AWS CDK), so a full environment came up in one step and matched the others instead of drifting, and automated the release path onto Lambda, Fargate, and S3.

Brigham Young University

• Built and taught a data science course spanning both halves of the work: the statistics (regression, time series, survival analysis) and the engineering to put a model somewhere a business can actually use it (Python, Linux, MySQL, Git, containers, testing, deploying to AWS).
• Advised statistics and information-systems students analyzing real data for local businesses, where the problem arrived vague and the data arrived messy, and coached them to present findings to the client in terms the client could act on.

Progrexion

• Built the survival-analysis model behind subscription retention: rather than just flagging who might cancel, it predicted the hazard of cancellation over time, so retention got a ranked list with timing attached and the business got a defensible turnover number instead of an assumption.
• Built the gradient-boosting model scoring how likely each negative item on a client's credit report was to be successfully removed, and put it in production to order every client's work queue by expected payoff, so effort went to the items that would actually change their report.
• Ran the A/B tests that told the business what a change was actually worth before it rolled out company-wide, fielded the ad hoc analyses leadership asked for when a number moved and nobody knew why, and built the metric the company used to gauge employee performance.

Education

Utah State University

Utah State University

Skills

• Software architecture
• AWS
• Kubernetes
• Distributed computing
• Python
• SQL
• Machine learning
• Statistics
• Data engineering
• Infrastructure as code
• CI/CD
• Observability & SLOs
• Incident response
• Cost optimization