I engineer high-throughput distributed systems, database storage engines, and cloud platforms with first-principles mechanical sympathy — from Linux kernel syscalls (`epoll`, zero-copy `sendfile()`) and memory barriers up to Raft consensus, LSM-tree compactions, and event-driven architectures.
A production-grade, 13-week curriculum covering 800+ technical modules. Synthesizes canonical textbooks (DDIA, OSTEP, Effective Java, Database Internals), university courses (MIT 6.5840, Stanford CS244B), and 28 real-world Alex Xu high-level design systems.
A high-performance embedded key-value storage engine engineered in Rust. Features Write-Ahead Logging (WAL) for durability, MemTable SkipList for lock-free concurrent ingestion, binary SSTable layout with Bloom Filters, and background leveled compaction.
High-concurrency, coordinate-free unique 64-bit ID generator based on the Twitter Snowflake specification. Eliminates network coordination round-trips while guaranteeing time-sortable IDs across multi-datacenter clusters.
End-to-end event streaming architecture engineered with Apache Kafka, Zero-Copy Linux PageCache socket transfers, and partitioned consumer group rebalances. Incorporates idempotent producers and transactional outbox patterns.
Complete implementations of classic staff-level LLD systems: thread-safe LRU Cache with MPSC write buffers, concurrent Elevator Dispatching Engine, Multi-Floor Parking Lot, and distributed Token Bucket Rate Limiters.
An architectural breakdown of read, write, and space amplification. Explains why LSM-Trees dominate write-intensive workloads through sequential append-only logs and compaction, while B+ Trees excel at point lookups via slotted-page layouts.
Deep dive into Project Loom continuation mechanics, carrier thread unmounting, and why `synchronized` blocks and JNI calls cause carrier thread pinning. Shows migration patterns to `ReentrantLock` and `StructuredTaskScope`.
How to architect multi-region rate limiters capable of handling 100,000+ RPS. Compares Redis Lua script token buckets against sliding window approximations, addressing race conditions and clock drift.
Tracing system calls from disk to socket. Explains how `epoll_wait` uses red-black trees and ready lists to achieve O(1) scaling, and how `sendfile()` eliminates user-space context switches through DMA copy.
Complete high-level design walkthrough for TinyURL. Features base62 encoding calculations, Key Generation Service (KGS) pre-allocation, 80/20 caching distribution, and partition strategy for 5-year storage scaling.
Understanding why Raft is understandable without sacrificing rigor. Detailed walkthrough of the Election Safety, Leader Append-Only, and State Machine Safety properties.
Low-latency engineering observations, gotchas, and architectural trade-offs. Join the discussion and reply.
synchronized or JNI that pins carrier threads, your pool deadlocks. For 1M idle WebSockets, Netty's epoll event-loop wins on memory efficiency.
rocksdb.write-stall-micros!
12 foundational computer science texts distilled into architectural mental models and flashcards.
Software engineer specializing in distributed systems, high-throughput JVM applications, storage engines, and cloud platforms. Proven track record of designing resilient systems processing hundreds of thousands of events per second with sub-15ms latencies.
I am passionate about building fault-tolerant distributed platforms and sharing deeply researched engineering knowledge. This personal hub unites my production projects, architectural essays, and the comprehensive 2026 Master Study Repository.