[ PROFILE / 01 ]

Hi, I'm
Anant Mishra.

I build production AI systems and the full-stack apps around them. Most of my work lives at the intersection of LLM agents, retrieval pipelines and scalable backends — the parts of an AI product that decide whether it actually works for real users.

Lately I've been deep in multi-agent RAG platforms with avatar interfaces — the kind of system where retrieval design, streaming and observability matter more than the model itself. Before that, motion-driven marketing sites tuned for speed and SEO. On the side I publish open source on NPM and ship MCP servers for creative tools.

I care about three things: latency, observability, and not lying to users. If your product is going to talk to humans, the words it says had better be grounded in your data and shipped fast enough to feel alive.

// Facts
Based
Noida, IN
Focus
AI agents & RAG systems
CodeChef
1400 rating
LeetCode
200+ solved
Open Source
lightweight-expression-detector / 109+ wkly installs
[ PRINCIPLES / 02 ]
P01

Ship the boring path first

Most AI demos break on edge cases nobody planned for. I build the unhappy paths before adding agents.

P02

Filter, don't trust

RAG is 90% retrieval design — chunking, metadata, filters. The model is the easy part.

P03

Streaming UX or it didn't happen

If the user is staring at a spinner for an LLM call, it feels broken. I default to streamed responses.

P04

Logs > guesses

Sentry, structured logs, tracing on every agent call. Production AI without observability is theater.

[ OPEN SOURCE / 05 ]

Open.source

Published & free to use
12 / 2025
Open Source / Dev Tools

ctxrelay

Stop re-explaining your project every time you switch AI agents.

A zero-dependency CLI that maintains a git-diffable .aicontext/ folder in your repo — a compact, token-budgeted map of your codebase and project state. Switch between Claude Code, Codex, Cursor or a fresh session and the agent picks up where the last one left off, without re-reading the whole tree.

TypeScriptNode.jsWASM tree-sitterCLI
05 / 2025
Open Source / NPM

Lightweight Expression Detector

Real-time facial expression detection. Zero dependencies. 109+ weekly NPM installs.

A from-scratch JavaScript package that runs facial expression detection in browser and Node.js with no external deps. Targets builders who want emotion signals without dragging a 50MB ML library along.

JavaScriptTypeScriptComputer Vision
[ TOOLCHAIN / 04 ] // SYSTEM_LOAD

The Stack

CORE: 6 / OK
UPTIME: 99.97
/bin/zsh — anant@os● live
$
$./modules/lang.sh --list
[OK]
01 / Languages5 pkg
TypeScript
94
JavaScript
96
Python
80
C / C++
75
SQL
85
$./modules/fe.sh --list
[OK]
02 / Frontend5 pkg
Next.js
95
React
94
Tailwind
96
Framer Motion
82
shadcn/ui
88
$./modules/be.sh --list
[OK]
03 / Backend5 pkg
Node.js
92
Express
90
BullMQ + Redis
84
REST APIs
92
tRPC
70
$./modules/db.sh --list
[OK]
04 / Databases5 pkg
PostgreSQL
86
Supabase
92
MongoDB
78
Pinecone
90
Appwrite
75
$./modules/ai.sh --list
[OK]
05 / AI / ML6 pkg
OpenAI / GPT
92
Gemini
85
LangChain / Graph
88
RAG Pipelines
92
MCP
86
Vector Search
90
$./modules/ops.sh --list
[OK]
06 / DevOps & Infra6 pkg
AWS (S3 / EC2)
78
Docker
80
Linux
84
Git
95
Sentry
82
Vercel
92
agents.daemon: runningkernel: lang.runtime / v20.ltsshell: zsh 5.9 (x86_64)mem: freenet: connected
[ ENDPOINT ]

Want to build something?

Start a Project