Ngan Nguyen
San Jose
Summary
Staff software engineer with extensive experience in designing and scaling feature pipelines for feed ranking systems serving hundreds of millions of users. Proven ability to lead teams in implementing impactful solutions, including near-real-time pipelines that enhance content freshness and boost user engagement.
Overview
3
3
years of professional experience
Work History
Staff Software Engineer
LinkedIn
Sunnyvale
04.2026 - Current
- Collaborated with core AI team to launch generative recommender ranking for LinkedIn's main feed, serving 300 million monthly active users.
- Directed design and implementation of near-real-time member sequence generation pipeline, coordinating five engineers.
- Led data validation efforts, guiding engineers to resolve major system and data corruption issues during production ramp.
- Enhanced member sequence freshness from three days to 15 minutes, leading to a 0.5% increase in engaged feed sessions.
- Designed orthogonal reinforcement learning ranking system for feed optimization, boosting retention metrics.
- Worked with AI infrastructure and feed AI teams to implement first PyTorch GPU-CPU sequential modeling use case.
- Decoupled model training, publishing, and inference stacks for generative and reinforcement learning models, improving system modularity.
Senior Software Engineer
LinkedIn
Sunnyvale
10.2024 - 03.2026
- Led in-house integration of near-real time item embedding generation pipeline within two-tower EBR architecture, improving retrieval effectiveness.
- Collaborated with tech leads on data and AI platform to launch LinkedIn's first GPU-based ranking retrieval system, significantly increasing content freshness.
Software Engineer
LinkedIn
Sunnyvale
07.2023 - 09.2024
- Launched first engagement pipeline to production, decreasing CPU usage by ten times and memory usage by two times.
- Reduced write QPS pressure on downstream systems by 1000 times while improving feature freshness from hours to minutes.
- Developed engagement feature generation SDK library for Windows-based aggregation features, enhancing feature integration.
- Devised timer bucket-based sliding window algorithm to reduce computation and memory usage compared to Beam's native solution.
- Partnered with LinkedIn's stream processing team to address state corruption and faulty timer expiry, ensuring system reliability.
- Implemented fixes and rolled out changes for all engagement features powering LinkedIn's main feed ranking model.
Education
Bachelor of Science - Computer Science
University of Minnesota
Minneapolis, MN05-2023
Skills
- ML Feature Engineering, Near real-time signal processing, Big Data Processing, Apache Beam, Beam Flink, Apache Kafka, Spark, HDFS, online API, GRPC, data pre-processing in Modeling, Pytorch, LLM, RL
- Stream processing
Accomplishments
- Bhimani Family Scholar - Awarded to CS who excelled in Research and Academic Performance in Computer Science Apartment
- Dean's List
Research
- Peer Recommendation Interventions for Health-related Social Support: a Feasibility Assessment
Timeline
Staff Software Engineer
LinkedIn
04.2026 - Current
Senior Software Engineer
LinkedIn
10.2024 - 03.2026
Software Engineer
LinkedIn
07.2023 - 09.2024
Bachelor of Science - Computer Science
University of Minnesota