Alex Chi Z. (skyzh)
Alex Chi Z. (Chi Zhang, known online as skyzh) is an individual database-systems engineer, educator, and open-source creator—not an incorporated company. He builds and explains systems software, especially database storage, query processing, vector search, and LLM inference, through production work, teaching, and free hands-on courses.
At a Glance
- Systems and backend engineers
- Computer-science students and self-directed learners
- Database researchers and practitioners
- LLM-inference and serving engineers
- +2 more
AI Tools by Alex Chi Z. (skyzh)
(1)
Tiny-LLM
LLM Inference Course for Engineers
Discussions
No discussions yet
Be the first to start a discussion about Alex Chi Z. (skyzh)
Latest News
Published “Plan Representation: #1 Lesson Learned from Building an Optimizer.”
Tiny-LLM course work reached a major Week 4 coding-agent checkpoint, including a learner CLI and reviewed agent-loop material.
Mini-LSM released v202608 with a redesigned learning experience and new coding-agent implementation path.
Tiny-LLM restructured Week 2 into a cumulative learner-owned path and restored learner-owned optimization and speculative-decoding exercises.
Products & Services
A free hands-on Rust course in which learners build an LSM-tree storage engine, covering in-memory and on-disk formats, compaction, crash recovery, WAL, MVCC, garbage collection, and transactions; the 2026 edition also provides a guided coding-agent track.
A hands-on course for systems engineers to build an LLM-serving system with MLX and Qwen3 on Apple silicon, progressing from array operations through kernels, KV caching, batching, paging, scheduling, and a local coding agent. Co-created with Connor.
A Rust course that builds an in-memory vector database, SQL top-k integration through Apache Arrow/DataFusion, and exact and approximate indexes including IVFFlat, NSW, HNSW, and IVF-PQ.
A work-in-progress Rust course on typed scalar and array representations, nullability, generic adapters, runtime erasure, and vectorized expression evaluation.
Market Position
Chi's work is positioned as small, inspectable, build-it-yourself educational systems that connect theory to implementation. Mini-LSM complements rather than replaces production engines such as RocksDB, TiKV, and Neon; Tiny-LLM is explicitly framed as an LLM-serving counterpart to CMU's Needle project and offers a readable alternative to studying very large, highly optimized CUDA serving codebases; the vector course similarly teaches ANN algorithms and SQL integration rather than presenting only a packaged vector database.
Leadership
Founders
Alex Chi Z. (Chi Zhang / skyzh)
Software engineer on Databricks' Lakebase team; previously systems software engineer at Neon, database-system R&D intern at RisingWave Labs, storage-system intern at ByteDance, and storage intern at PingCAP. He holds an MS in Computer Science from Carnegie Mellon University and a BE in Computer Science and Technology from Shanghai Jiao Tong University.
Founding Story
This is a personal creator portfolio rather than a company with a separately documented incorporation or founding event. Chi says he enjoys solving difficult real-systems problems and explaining how they work; his educational projects turn that interest into small, readable systems that learners can build end to end. Tiny-LLM specifically began because optimized LLM-serving codebases were difficult to understand, so he chose to implement inference from scratch with array and matrix operations.
Business Model
Revenue Model
The published courses and source code are publicly available for free and receive ongoing updates and bug fixes. The Mini-LSM course states that paid code-review and office-hours services may also be offered; no subscription, API, advertising, or licensing revenue model is documented.
Target Markets
- Systems and backend engineers
- Computer-science students and self-directed learners
- Database researchers and practitioners
- LLM-inference and serving engineers
- Rust developers
- Educators teaching database systems or systems programming
- Learning database internals and storage-engine design
- Learning LLM inference and serving-system mechanics on an accessible Apple-silicon machine
- Learning vector-search algorithms and their SQL/query-planning integration
- Teaching database systems and providing course projects
- Experimenting with Rust systems programming, compilers, storage, query optimizers, and performance engineering
- Studying coding-agent workflows with bounded, validated, approval-based tool use