EXPERIENCE
Software Engineer (Python, Ray, K8s)
DoorDash, New York City, NY, July 2022 - present
- Promoted to senior software engineer on Machine Learning Platform team and led a new dedicated full stack forecasting team with SWEs and MLEs as techlead.
- Collabrorated with MLEs on mulitple projects from research prototypes to production:
- Built ML models with realtime features (Kafka + Redis) to detect delivery risks, integrated with multiple production services to run A/B experiments to measure quality and dollar impact.
- Worked cross teams to launch predictive scaling forecasts for K8s pods autoscaling based on QPS and target CPU utilizaiton, saving company compute infra 2M+/yr for 50+ micro-services.
- Designed distributed training pipelines with Ray Core to support a patent ensemble forecasting model, reducing 50% execution time and 90% cost for model training and batch inference.
- Developed a set of REST APIs and gRPC APIs for Forecast Platform backend to integrate with internal services and enabled 20+ user self-serve onboarding uses cases via Forecast UI.
- Set up GPU cluster infrastructures with KubeRay that empower MLEs fine tuned Llama 2 7B LLM in 30 mins per epoch, and set up Prometheus metrics, Tensorboard, Grafana, and Chronosphere dashboards for better ML Training Platform observability.
Software Development Engineer (Python, Spark, AWS)
Amazon, New York City, NY, March 2018 - July 2022
- Promoted from L4 to L5 in 2020 and supported a summer intern as engineer manager in 2022.
- Led end-to-end productionization of DL forecasting models for emerging Amazon businesses, including data preprocessing (Spark pipelines), model training (MxNet/Pytorch) and serving (SageMaker endpoints), and backtesting/evaluation automation (reduced from 7 days to 2).
- Collaborated closely with Data Scientists to transition research prototypes into production ML forecasts with continuous training and monitoring on AWS, which improved accuracy by 30% over baseline with upgraded model architecture, feature engineering, and hyperparameter optimization.
- Built a set of scalable real-time and batch ETL pipelines to ingest Wikipedia and news articles into Elasticsearch for Alexa QA. It sped up the batch loading with AWS Lambda from 4 hours to 20 minutes and total ingestion time from 10 to 2 hours.
Software Development Engineer Intern (Java, AWS)
Amazon Web Services, Seattle, WA, May - August 2017
- Implemented an exponential way to improve deployment speed by 50% with DynamoDB and SWF.
- Built a song quiz Alexa skill with flask-ask framwork and SSML output in Global Intern Hackathon.
EDUCATION
Carnegie Mellon University, Pittsburgh, PA & Silicon Valley, CA
Master of Science in Information Technology, Mobility, August 2016 - December 2017
Zhejiang University, Hangzhou, China
Exchange at Singapore Management University, Information Systems, August - December 2015
Bachelor of Engineering, Computer Science, September 2012 - July 2016
SKILL
Programming/Scripting Languages: Java, Kotlin, Python, Scala, C/C++, JavaScript, HTML, CSS, SQL
Dev env/AI adoption: VSCode + Copilot, IntelliJ + Augment, Cursor + ClaudeCode
MLOps/Infra Tools: K8s, SageMaker, ECR, Chronosphere, Dagster, Databricks, Snowflake, Kafka, Redis