Devakinandan K

Backend and applied AI engineer. Building reliable retrieval systems and developer products.

I build backend and applied AI systems with an emphasis on explicit contracts, inspectable behavior, and failure modes that are easy to reason about.

Recent work spans retrieval pipelines, real-time multiplayer products, and tools for capturing and sharing technical knowledge.

GitHub

Explore work & writing

Projects, practical notes, and longer engineering essays.

Projects

04

Shipped systems with the problem, ownership, stack, and evidence kept together.

Notes

02

Compact explanations of resolved questions, mental models, and practical gotchas.

Blogs

02

Longer arguments about engineering trade-offs and lessons from building.

Recent writing

Recent notes

View all
  1. What is Jev, and when should an agent use it?
    • agents
    • jev
    • inference
  2. Why does RAG need reranking after retrieval?
    • rag
    • retrieval

Recent blog posts

View all
  1. RAG quality is a pipeline problem, not a model choice
    • rag
    • retrieval
    • evaluation
  2. Reliable agent workflows need explicit state
    • agents
    • reliability
    • backend
View all projects

  1. Hybrid RAG

    A local workspace for uploading documents, exploring passages, and asking grounded questions with inspectable sources.

    Evidence

  2. Typeracter

    A focused typing app for solo practice and private real-time races with friends, with no account required.

    Evidence

  3. BugRace AI

    A real-time multiplayer debugging game with server-owned race state, semantic evaluation, and deterministic scoring.

    Evidence

  4. Second Brain

    A multi-user knowledge backend for saving links, extracting metadata, organizing notes, and sharing collections.

    Evidence