Sample document — fictional candidate for illustration purposes only. Your CV is prepared from your own record.
Professional Summary
Staff engineer with 10 years of experience across ML infrastructure, distributed systems, and large-scale platform architecture. At Meta, led the ground-up redesign of an ML inference pipeline now serving 80 million requests per day with 99.95% uptime. Owns three core platform services. Published researcher in distributed ML scheduling; PhD from Stanford. Known for technical clarity, cross-functional influence, and growing engineers into senior roles.
Tech Stack
Python
C++
Kubernetes
Apache Spark
PyTorch
Thrift / gRPC
Apache Flink
Presto
Hive
HDFS
Grafana
Scuba
Experience
Meta, London
- Architected and led delivery of the next-generation ML inference gateway — a Python/C++ hybrid serving 80M req/day with p99 latency under 12 ms; zero-downtime rollout over 6 months.
- Own three platform services (model registry, feature serving cache, batch scoring scheduler) across a combined on-call surface of 40+ engineers.
- Drove adoption of Kubernetes-based autoscaling for inference workers, reducing over-provisioning cost by £2.1M/year across EMEA serving clusters.
- Mentored 4 senior engineers; 2 promoted to staff level within 18 months under structured technical growth plans.
- Authored two internal RFCs — on model versioning semantics and canary evaluation standards — adopted as org-wide policy in 2024.
Twitter, San Francisco
- Built Spark-based pipeline to compute user engagement signals at scale; reduced feature freshness lag from 4 hours to 18 minutes.
- Designed schema evolution framework for Thrift-based event logs, enabling zero-downtime migrations across 200+ downstream consumers.
- Led a 5-person squad delivering real-time ad targeting feature store; contributed to 6% improvement in ad relevance scores.
Google DeepMind (intern → researcher), London
- Developed distributed task scheduler prototype in C++ for large-scale RL training jobs; results informed internal job orchestration tooling.
- Co-authored conference paper on adaptive resource allocation for ML workloads (NeurIPS 2016 workshop).
Education
PhD Computer Science — Distributed Systems
Sep 2012 – Jun 2015
Stanford University — Advisor: Prof. John Ousterhout; Thesis on fault-tolerant ML scheduling
BEng Software Engineering
Sep 2008 – Jun 2012
Tsinghua University, Beijing — Ranked 3rd in department, summa cum laude
Selected Publications & Talks
NeurIPS 2016 Workshop — "Adaptive Scheduling for Distributed ML Training" · QCon London 2023 — "Lessons from 80M req/day: ML Serving at Meta Scale" · 3 internal Meta tech talks with 400+ engineering attendees