● Project Sequence

CSE Projects Track

A curated list of six projects to help CSE students build and test networking, search, storage, and distributed systems.

STAGE 00 Prerequisite

HTTP Server & Storage Latency

Build an HTTP service with relational database persistence. Measure and compare p50 and p99 latency between memory lookups and disk database transactions under load.

HTTP/1.1 Protocol Relational Persistence Connection Pooling wrk Latency Profiling

Goal: Understand client-server request lifecycle and storage bottlenecks. View Specification →

STAGE 01 Networking & Protocols

In-Memory Key-Value Store (Redis Clone)

Implement a TCP socket server that speaks the binary RESP wire format. Handle concurrent connections using non-blocking I/O multiplexing, key expiration, and append-only durability.

TCP Sockets Binary Wire Parsing epoll / Event Loop Active/Passive Eviction AOF Durability

Goal: Master byte streams, event loops, and low-level network servers. View Specification →

STAGE 02 High-Dimensional Algorithms

Approximate Nearest Neighbor (ANN) Vector Index

Implement vector distance functions (Cosine, Euclidean). Build an exact flat scan baseline, then an Inverted File (IVF) index partitioned with k-means centroids. Measure Recall@10 versus QPS.

Vector Distance Math Priority Heaps k-Means Clustering Inverted File Indexing Recall@10 Evaluation

Goal: Understand high-dimensional search and vector database internals. View Specification →

STAGE 03 Storage & Hashing

Content-Addressable Object Store (Git Internals)

Build a content-addressable storage engine using SHA-1 hashing and zlib stream compression. Construct blob, tree, and commit DAG objects readable by the official git binary.

Content Addressing SHA-1 Hashing zlib Deflate/Inflate Binary Tree Packing Commit DAG Traversal

Goal: Master immutable storage, cryptographic graphs, and file formats. View Specification →

STAGE 04 Storage Engines

LSM-Tree Disk Storage Engine

Build an embedded persistent key-value engine with Write-Ahead Logging (WAL), an in-memory SkipList MemTable, immutable SSTable disk files with binary search footers, and Bloom filters.

Write-Ahead Logging SkipList MemTable SSTable Binary Files Bloom Filters Compaction Iterators

Goal: Master database storage engines and crash-safe persistence. View Specification →

STAGE 05 Distributed Systems

Distributed Consensus Engine (Raft Protocol)

Implement a replicated state machine across 3 nodes using Raft. Handle randomized election timers, leader transitions, heartbeat broadcasts, log replication, and partition split-brain recovery.

State Machine Replication Leader Election Quorum Majority Arithmetic Log Invariant Matching Split-Brain Fault Tolerance

Goal: Master distributed fault-tolerance and consensus protocols. View Specification →