Reading
Papers & Books
Technical literature I've read, am reading, or plan to. Mostly distributed systems, databases, data engineering, and AI. Notes are my own.
Distributed Systems
The paper that defined large-scale batch processing. The simplicity of the programming model is the insight.
Foundational for understanding wide-column stores and the architecture of HBase, Cassandra.
Eventual consistency, vector clocks, consistent hashing. Still the best paper on availability-first design.
TrueTime and externally-consistent distributed transactions. Remarkable engineering.
Much easier to understand than Paxos. Good foundation for understanding etcd, CockroachDB.
Still not that simple, but essential reading. Read Raft first.
The best systems book written in the last decade. Required reading for anyone building data systems.
Databases
Understand MVCC and B-tree vs LSM-tree tradeoffs. Explains much of MongoDB's storage behavior.
Good read on how a production storage engine evolves under real workloads.
Architecture of how Pinot handles real-time + offline data for low-latency analytics at scale.
Best resource for understanding database indexing across engines. Practical and deep.
Data Engineering
The original Kafka paper. The log abstraction is the key insight. Still holds up.
The conceptual foundation for Flink, Beam, and modern stream processing. Windowing, triggers, accumulation.
ACID on top of Parquet in object storage. Foundational for understanding modern lakehouse architectures.
The original Spark paper. RDD model and in-memory computation for iterative algorithms.
AI & Machine Learning
The transformer architecture paper. Everything in modern AI traces back here.
MLA, DeepSeekMoE, FP8 training. Remarkable efficiency. Changed the conversation on inference cost.
Good detail on training data curation, instruction tuning, and safety filtering at scale.
Sparse mixture-of-experts applied to an open model. MoE becomes practical.
The CoT paper. Step-by-step reasoning significantly improves complex task performance.