# Kevin Zhu

Kevin Zhu is a remote worker on Himalayas, Machine Learning Engineer at Reddit, based in United States.

Source: https://himalayas.app/@kevinzhu

```text
Machine Learning Engineer at Reddit building Ads retrieval platforms, including Keyword Targeting that increased eligible impression matches by 18%.
```

| Detail | Value |
| --- | --- |
| Availability | Open to opportunities |
| Location | United States |
| Authorized to work in | United States |
| Job categories | Machine Learning Engineer, Software Developer |
| Skills | Machine Learning, Production ML, Deep Learning, Distributed Training, Real Time Inference, MLOps, Recommendation Systems, Ads Machine Learning, Candidate Retrieval, Learning To Rank, ANN Search, Fraud Detection, Demand Forecasting, Smart Bidding, Conversion Modeling |
| LinkedIn | https://linkedin.com/in/kevin-zhu-24210370 |

## About

```text
At Reddit, I architected a shared PyTorch two-tower retrieval platform for Ads, supporting distributed GPU training, embedding generation, and real-time vector retrieval for Targeting and Shopping Ads. I scaled Keyword Targeting to general availability, increasing the share of eligible impressions matched to advertiser keywords by 18%.

I improved training and retrieval performance at Reddit, reducing model training time by 25% and P95 retrieval latency by 20%. I also built an LLM-based evaluation agent that reduced manual evaluation effort by 40%.

At Rokt, I developed conversion modeling and Smart Bidding systems, improving CVR model log loss by 5% and reducing advertiser cost per acquisition by 8% in online campaign experiments. At Meta, I worked on Facebook Dating Anti-Scam Detection, increasing scam detection recall by 12% at a fixed false-positive rate.

At Amazon, I developed ML components for Alexa speech recognition and natural language understanding, and built distributed demand forecasting pipelines for Amazon Retail. My work improved slot-level F1 by 3 percentage points and reduced forecasting data preparation time by 22%.
```

## Experience

### Machine Learning Engineer (current)

Reddit. Sep 2024 - Present

```text
Architected Reddit Ads' shared PyTorch two-tower retrieval platform, enabling distributed GPU training and real-time vector retrieval. Scaled multi-GPU training with PyTorch DDP and Kubeflow, reducing training time by 25% and improving retrieval latency by 20%.
```

Skills: PyTorch, Distributed-Training-(PyTorch-DDP), Kubeflow, Two-Tower-Neural-Network, ANN-Search, Faiss, Milvus, HNSW, Semantic-Search, Keyword-Targeting, LLM-Evaluation, Distributed-Training, Model-Deployment, CI-CD, Docker, Kubernetes

### Machine Learning Engineer

[Rokt](https://himalayas.app/companies/rokt). Apr 2023 - Aug 2024

```text
Developed production ML systems for Ads Conversion Modeling and Smart Bidding, combining TensorFlow-based conversion prediction and real-time inference. Improved CVR model performance and reduced advertiser cost per acquisition by 8% in online experiments.
```

Skills: TensorFlow, Feature-Engineering, Real-Time-Inference, AWS, A-B-Testing, Probability-Calibration, Smart-Bidding, CVR, Delayed-Conversion-Modeling

### Machine Learning Engineer

[Meta](https://himalayas.app/companies/meta-inc). Mar 2021 - Feb 2023

```text
Developed ML models for Facebook Dating Anti-Scam Detection using PyTorch and graph-based features. Increased scam detection recall by 12% at a fixed false-positive rate and reduced P95 scoring latency by 17%.
```

Skills: PyTorch, FBLearner, Hive, Presto, Feature-Engineering, Class-Imbalance-Handling, Prediction-Threshold-Optimization, Graph-Based-Features, Behavioral-Classification, Model-Inference

### Software Development Engineer

[Amazon](https://himalayas.app/companies/amazon). Jul 2019 - Jan 2021

```text
Developed ML components for Alexa AI's speech recognition and natural language understanding systems. Improved intent classification and slot extraction, increasing slot-level F1 by 3 percentage points.
```

Skills: Machine-Learning, Speech-Recognition, Natural-Language-Understanding, Sequence-Modeling, Neural-Networks, Backend-Integrations, Python, Java, C++

## Education

### Cornell University

Master's Degree, Computer Science. 2015 - 2016

```text
Master's Degree in Computer Science from Cornell University, completed in 2016.
```

### University of Wisconsin-Madison

Bachelor of Science, Electrical Engineering. 2008 - 2012

```text
Bachelor of Science in Electrical Engineering from the University of Wisconsin-Madison, completed in 2012.
```

## Tech stack

- Apache Spark
- Presto
- Amazon EC2
- Amazon S3
- Kubernetes
- NumPy
- Pandas
- Databricks
- Python
- Java
- C++
- Scala
- TensorFlow
- PyTorch
- scikit-learn
- Keras
- Kubeflow
- Milvus
- Docker
- SQL
- XGBoost
- LightGBM
- Faiss
- AWS

To contact Kevin, message them from this profile with a Himalayas employer account: [Hire on Himalayas](https://himalayas.app/hire).

## People also viewed

- [Martin Chavez User](https://himalayas.app/@martinchavezuser1): Principal ML Engineer, TechHolding
- [kunal narang](https://himalayas.app/@kunalnarang): Senior AI / ML Engineer, Dentsply Sirona
- [Michael Li](https://himalayas.app/@michaelli): Senior Machine Learning Engineer, Snap Inc.
- [Dennis Scott](https://himalayas.app/@dennisscott): Senior Machine Learning Engineer, Attentive
- [Deshna Vemula](https://himalayas.app/@deshnavemula): AI/ML Engineer, Amazon Ads
- [Alex Xiao](https://himalayas.app/@alexxiao): Senior Machine Learning Scientist, Tesla
- [Kevin Shi](https://himalayas.app/@kevinshi1): Senior Machine Learning Engineer, Netflix
- [Yongwoo Noh](https://himalayas.app/@yongwoonoh): Senior Machine Learning Engineer, Pinterest
- [Matthew Magsombol](https://himalayas.app/@matthewmagsombol): Senior Software Engineer, Machine Learning, Reddit, Inc.
- [Terry Vickery](https://himalayas.app/@terryvickery): Senior Machine Learning Engineer, Reddit, Inc.
