Arjun IyerEngineering & product
← Selected work

Learning & developer tools / Live product + open-source edition

l33.bot
Turn interview preparation into deliberate practice.

Coding-interview practice with a database-backed problem catalogue, curricula and learner progress. An open-source edition supports configurable language-model providers.

The product problem

A practice product needs durable problems and learner progress alongside model assistance. The hosted service and a self-hosted edition have different identity and payment needs.

Engineering & product decisions

Persist the learning structure

PostgreSQL backs the problem catalogue, curricula and progress so the experience has structure beyond individual prompts.

Define a provider boundary

The open-source edition exposes a configurable provider layer for OpenAI, Anthropic and Google.

Keep the editions distinct

The hosted app has account and payment integration. The open-source edition intentionally excludes auth and payments.

Make production routing verifiable

Deployment work includes callback origin, return-path and sign-out regression checks, preserving existing database and session boundaries.

What the evidence supports

The hosted route and public source edition are available. A reachable product and regression checks do not establish learner outcomes or traction. Provider access and keys are configured by the self-hosting user.

This case study describes Arjun Iyer’s product and engineering work through Fallible Software, Co. Model services, frameworks and upstream runtimes retain their own provenance. Evidence and status notes →

Discuss this kind of work ↗