Zheren Dong

AI Research Engineer

Building LLM training, retrieval, and data systems for coding agents.

Zheren Dong

I’m a research engineer on the AI Research Team at Augment Code, where I focus on post-training LLMs and data curation strategies that improve retrieval performance for coding agents in production.

My work includes training-data pipelines, distribution shifts, model onboarding, and evaluation. Outside of work, I pursue independent ML research; my recent work explores how spelling-aware embeddings can improve language modeling across benchmarks.

Previously, I worked at Applied Intuition and Rivos (now part of Meta).

News

Jan 25, 2026 New preprint on Spelling Bee Embeddings for Language Modeling is now on arXiv! 🎉

Publication

Illustration for Spelling Bee Embeddings

Spelling Bee Embeddings for Language Modeling

Markus N Rabe, Judith Clymo, and Zheren Dong
arXiv preprint arXiv:2601.18030, Jan 2026
We add spelling information to token embeddings. Across models with 40M to 800M parameters, the approach improves spelling and standard benchmarks; scaling results suggest an equivalent of about 8% less compute and data for the same test loss.
Read on arXiv

Experience

Jan 2025 – PresentPalo Alto, CA

Member of Technical Staff, AI Research Team

Augment Code
  • Retrieval performance and context engineering
  • Embedding model training and data curation
  • Model onboarding and evaluation
Sep 2023 – Jan 2025Mountain View, CA

Software Engineer, Vehicle Platform Team

Applied Intuition
  • Next-gen Software Defined Vehicle (SDV) platform
  • Data infrastructure for vehicle telemetry and fleet health monitoring
  • On-board runtime environment and applications
Jun 2022 – Aug 2023Mountain View, CA

Member of Technical Staff

Rivos Inc.
  • Rust runtime support library for RISC-V system bootstrapping
  • Rust-based DDR5 SPD decoder/encoder CLI tool per JEDEC standard (intern project in summer 2022)
May 2021 – Sep 2021Beijing, China

Software Engineer Intern

Alibaba Group
  • Redesigned TensorFlow-based user vector generation module in C++ for vector and tree-based deep match retrieval system

Education

2021 – 2022Irvine, CA

M.S. Computer Science

University of California, Irvine
2016 – 2020Santa Barbara, CA

B.S. Computer Science (Honors)

University of California, Santa Barbara

Skills

Languages

Python · C/C++ · Rust · Go · Java · TypeScript · SQL

ML & Data

PyTorch · TensorFlow · Ray · Spark · CUDA

Infrastructure

Kubernetes · Docker · GCP · AWS · BigTable · BigQuery · Kafka · Redis · PostgreSQL