← Back to Data & Analytics All sample CVs →

Sample document — fictional candidate for illustration purposes only. Your CV is prepared from your own record.

Luke Zhang
Machine Learning Engineer
luke.zhang@email.com +65 9123 4567 Singapore linkedin.com/in/lukezhang github.com/lukezhang-ml
Professional Summary

ML Engineer with 4 years of experience building and deploying large-scale recommendation and personalisation systems at ByteDance Singapore. Specialises in the full model lifecycle — from offline feature engineering and training on Spark to real-time serving at low latency with Redis and Kubernetes. Proven track record shipping models that move engagement metrics for tens of millions of daily active users.

Impact Highlights
14 M Daily active users served by models
+22% Click-through rate lift, reco v3
38 ms p99 inference latency, prod serving
4 TB/day Feature pipeline throughput
Technology Stack
PyTorch Python Spark Redis Kubernetes Docker Kafka Triton Inference Server MLflow Airflow HDFS Hive Git Helm
Experience
Senior Machine Learning Engineer Mar 2024 – Present
ByteDance, Singapore
  • Business question: Could a two-tower retrieval model replace heuristic candidate generation and improve downstream ranking quality? Approach: Designed and trained a FAISS-backed two-tower model on 90 days of interaction history; shadow-tested against the heuristic for 3 weeks. Outcome: Adopted in production — recall@100 improved 18%, contributing to a 22% CTR lift on the recommendations surface.
  • Business question: Was GPU over-provisioning inflating serving costs? Approach: Profiled inference jobs, identified 40% idle GPU time, migrated batching logic to Triton dynamic batching with auto-scaling on Kubernetes. Outcome: Serving cost reduced 29% while p99 latency fell from 72 ms to 38 ms.
  • Lead engineer for the feature store migration from Redis Cluster v5 to v7; coordinated zero-downtime cutover for 14M DAU traffic.
Machine Learning Engineer Jul 2022 – Feb 2024
ByteDance, Singapore
  • Business question: How could we reduce model training time to enable daily retraining at scale? Approach: Rewrote the Spark feature pipeline to use delta tables with incremental processing, cutting full-scan jobs. Outcome: Training cycle shortened from 11 hours to 3.5 hours, enabling same-day model freshness.
  • Business question: Was the ranking model underperforming for new users with sparse history? Approach: Added a cold-start branch using content-based embeddings blended with collaborative signals via a gating network. Outcome: New-user 7-day retention improved 9 percentage points in A/B test across 620,000 users.
  • Built and maintained MLflow experiment tracking infrastructure; standardised model cards for 12 production models.
Data Scientist (Graduate) Aug 2021 – Jun 2022
ByteDance, Singapore
  • Business question: Could watch-time prediction be improved over the existing click-based model? Approach: Explored watch-time regression with position-bias correction on logged data; ran offline evaluation using NDCG@10. Outcome: Model selected for A/B test; average session length increased 7% over 4-week test.
  • Authored internal wiki covering feature engineering best practices adopted by 3 squads.
Additional Skills
  • Recommender systems design
  • A/B testing & experimentation
  • Distributed training (DDP)
  • Feature store architecture
  • System design interviews
  • English (fluent), Mandarin (native)
Education
BSc Computer Science Sep 2017 – May 2021
Hong Kong University of Science and Technology (HKUST) — First Class Honours, Dean's List 2019–2021