$ echo "Hello, world. I'm"
// Software Engineer
Backend engineer at DoorDash, working on Ads & Promotions. I build event-driven systems on Kafka, distributed rate limiting on Redis + Cassandra, and the CI/CD guardrails that keep them honest at production scale.
I'm John — backend engineer on the Ads & Promotions team at DoorDash. I spend my days decoupling synchronous code paths into Kafka pipelines, designing rate-limiting frameworks on top of Redis and Cassandra, and pushing deployments through Spinnaker canaries with automated regression detection.
Before that I interned twice at DoorDash on Supply & Demand and Consumer Android, and graduated from Connecticut College with high honors in CS and Statistics & Data Science. I care about idempotency, low-latency tail behavior, and writing the kind of services on-call engineers don't dread.
$ cat whoami.json { "name": "John Mwai", "role": "Software Engineer @ DoorDash", "team": "Ads & Promotions", "focus": ["kafka", "rate-limiting", "k8s"], "langs": ["java", "kotlin", "python", "go"], "based": "new york, ny", "alma": "connecticut college '23" } $ uptime since aug 2023 · 6+ infra projects shipped · 2 mentees
$ git log --author="john" --oneline
DoorDash · New York, NY
Owning event-driven infrastructure for promotion endpoints serving production traffic. Shipped 6+ complex projects spanning Kafka pipelines, distributed rate limiting, multi-cell scheduling, zero-downtime migrations, and CI/CD guardrails. Mentor 2 junior engineers on distributed-systems patterns and interview candidates for the team.
DoorDash · San Francisco, CA
Designed a hotspot prediction system using DBSCAN clustering over hybrid real-time and historical geospatial data for dynamic Dasher allocation. Built supporting Kotlin pipelines and Python batch jobs to refresh clusters at city scale.
DoorDash · San Francisco, CA
Redesigned the in-app rating flow and shipped it through A/B testing with statistical significance analysis. Incremental rollout to 10M+ users delivered a 4× engagement increase.
Connecticut College · New London, CT
GPA 3.85 / 4.00 (Major 3.93). Dean's High Honors, Dean's Scholarship, CODE2040 Fellow. Coursework in distributed systems, operating systems, databases, networks, algorithms, linear algebra and probability.
A short list of infra I've shipped end-to-end at DoorDash — pipelines, frameworks, and the load-bearing middle.
Decoupled synchronous Ads & Promotions code paths into a Publisher → Topic → Consumer pipeline with gRPC cross-service calls. Consumer-group autoscaling, idempotent processing with at-least-once guarantees, dead-letter queues, and partition-based parallelism for horizontal scale.
Token Bucket and Sliding Window algorithms with LOCAL, HYBRID, and DISTRIBUTED strategies. Redis for distributed caching, Cassandra for persistent state, automatic circuit breaking and exponential backoff to keep partial failures from cascading.
Replaced a single-pod scheduler with a distributed coordinator using consistent hashing for job distribution, cell-aware filtering, and environment-based routing. 4× capacity increase with graceful failover across data centers.
Led a zero-downtime cluster deprecation across six microservices using a dual-write migration pattern, coordinated schema evolution, and automated consistency verification. Eliminated a single point of failure while keeping data integrity intact end-to-end.
CI/CD automation with Spinnaker canary analysis and statistical anomaly detection across six high-traffic promotion endpoints. Automated regression detection plus rollback hooks that catch faulty releases before they ever reach 100% traffic.
City-scale geospatial clustering for Dasher allocation: DBSCAN over a hybrid of real-time and historical signals to forecast high-density order regions. Production data pipelines in Kotlin with SQL analytics and Python batch jobs to refresh clusters.
$ ls -la projects/ → 6+ shipped, more behind NDAs.
Always happy to chat about distributed systems, event-driven architecture, or career advice for new grads heading into infra. Email is the fastest way through.