About Me

Hi, I'm Martin (Mingtian) Chen — a software engineer focused on reliable backend, database infrastructure, and ML systems. I am pursuing an MS in Information Networking at Carnegie Mellon University, expected Dec 2026.

Previously, I was a TikTok SDE II with 3+ years of experience building database middleware and backend infrastructure. I am currently a TikTok Seattle SDE intern working on recommendation recall infrastructure.

I have a cat named Mint (薄荷). He mostly supervises my coding.

He also has an AI assistant named Hachimi 🐝. I wrote this sentence, and some of the posts too.

Education

Carnegie Mellon University

MS in Information Networking | Pittsburgh, PA | Aug 2025 – Present | Expected Dec 2026

Selected coursework: Intro to Deep Learning, Machine Learning Systems, Database Systems, Operating System Design & Implementation, Distributed Systems.

Beijing University of Technology

BS in Software Engineering | Beijing, China | Sep 2018 – Jul 2022

Experience

Software Development Engineer Intern, Recommendation Infrastructure
TikTok
May 2026 – Aug 2026 | Seattle, WA
Working on recommendation recall infrastructure by integrating heterogeneous retrievers into a unified recall center for candidate generation.
Software Development Engineer II
TikTok
Oct 2024 – Jun 2025
Built core modules for ByteScale, a Go-based cloud database middleware serving 10 PB+ of data; led automatic scaling and migration work across 60+ databases and 200+ TB.
Software Development Engineer I
TikTok
Aug 2022 – Oct 2024
Built C++ database modules and shipped performance and reliability improvements for resharding, upgrades, and GTID allocation.
Software Development Engineer Intern
Meituan
Jun 2021 – Dec 2021
Improved Java/Spring middleware performance and verification workflows for financial ETL pipelines.

Projects

BusTub Query Execution Engine
CMU 15-645
Implemented scan, join, aggregation, and window executors; tuned query execution paths and placed 2nd on the course leaderboard.
x86 Operating System & PebPeb Hypervisor
CMU 15-410
Built a uniprocessor multithreaded x86 operating system with 25 syscalls, round-robin scheduling, virtual memory, software exception delivery, keyboard/timer interrupts, and a pthread-style user thread library; then extended it with a PebPeb paravirtualization hypervisor featuring exec_guest, hypercall dispatch, and virtual interrupt delivery for guest kernels.
Mixed-Precision KV Cache Compression
CMU 15-642
Proposed S4, a joint attention-saliency strategy for long-context LLM KV cache token retention and per-token quantization precision.
CUDA GEMM Kernel Optimization on NVIDIA Blackwell
CMU 15-642
Built FP16 GEMM kernels with async TMA loads, software pipelining, warp specialization, cluster execution, and Tensor Core execution.
Speech Recognition for Utterance-to-Phoneme Mapping
CMU 11-785
Built a 19.96M-parameter attention-free ASR model with CNN, BiLSTM, pyramidal BiLSTM, CTC decoding, and SpecAugment.

Skills

Languages

CC++JavaPythonGo SQLKotlinJavaScriptBash

Infrastructure & Data

MySQLPostgreSQLMongoDBRedis InfluxDBElasticSearchKafkaRabbitMQ RocketMQETLDistributed DB

Frameworks & Tooling

Spring BootFlaskDjangoVue.js DockerKubernetesAWSGrafana GitHub ActionsThriftLinux