Shamsher Singh
Java and Guidewire · Axum · OpenAI streaming · persistent documents
“The project gave my Java background a real Rust architecture to explain and defend.”
Shamsher brought enterprise software experience from Java and Guidewire. The engagement focused that background into Crust AI, a Rust chat platform with authentication, persistent documents, OpenAI streaming, and a backend architecture that can be discussed line by line.
The goal was not to hide behind an AI API call. It was to build the boundaries around that call: who can access chat, how replies stream, where documents live, and which constraints keep the system predictable.
Enterprise experience, reframed as Rust backend ownership
Shamsher came in with Java and Guidewire experience. We used that domain depth as an advantage, then moved the proof into an Axum application with explicit users, sessions, routes, and protected boundaries.
- ✓Axum backend with a clear route and domain boundary
- ✓User and session flow designed as application behavior, not a tutorial stub
- ✓Rust ownership and async concepts attached to a system he could explain
The chat starts behind an auth gate
User
email + password
JWT cookie
session boundary
Protected chat
user-owned data
Postgres
users, sessions, chats
Reply limiter
30 replies / hour / user
Streaming AI replies without hiding the hard parts
The interesting system behavior sits between a user prompt and a streamed assistant reply. The backend owns the OpenAI call, forwards deltas through SSE, and keeps the response bounded and observable.
- ✓Prompt enters through an authenticated chat boundary
- ✓OpenAI output becomes a controlled stream of delta events
- ✓Partial replies can be stopped without losing the system state
The reply arrives as a live stream
Prompt
persisted user message
OpenAI
token stream
SSE
delta events
The frontend can abort the stream, persist the partial assistant reply, retry a failed send, and keep chat history consistent.
Documents and chat context make the project feel real
Crust AI is more than a streaming demo. Persistent documents, recent messages, upload limits, and explicit document lifecycle create a system with real product constraints to reason about.
- ✓Documents stored and attached to future chat context
- ✓Bounded input surface with explicit upload limits
- ✓Remove and attach behavior modeled as deliberate domain actions
Files become chat context before the request
Drop file
text or image
Postgres
draft document
Chat context
recent docs + messages
Bounded input
5 MB upload limit
Remove or attach
explicit document lifecycle
The journey
One clear path through the build
Scope
Crust AI defined as a backend-first Rust chat platform with a real domain model.
Auth
Axum routes, users, sessions, and protected chat boundaries wired together.
Streaming
OpenAI token output streamed through the backend with bounded reply behavior.
Documents
Persistent document lifecycle added so chat context survives beyond one request.
Now
A focused Rust AI system that turns enterprise experience into backend proof.
Drag to explore · Select a milestone to read the full story
Mentor perspective
Shamsher did not need another generic CRUD tutorial. He needed a project where authentication, streaming, persistence, and product boundaries forced the Rust concepts to become decisions. Crust AI gave him that system to build and defend.
Max Wells
Want a Rust AI backend story like Shamsher's?
Available for callsMax Wells
Founder · Rustify
Book your Rust career diagnostic
30 minutes, 1:1 with Max. An honest read on where you are and whether Bootcamp or Mentorship fits, before you apply.
Only a few diagnostic slots left this week.
Applications are reviewed carefully.
More client journeys
