Skip to content
Arjun MDeveloper & explainer
Arjun M / Index
Backend systems. Human explanations.
AI agent automation × backend systems

Building things.
Making sense
of them.

Explore the work
A little context

I build backend tools and explain how to use them, so a good idea can become a working integration.

Systems that work. Stories that make sense.Scroll to explore
The through line

I like working across the technical work and the explanation around it.

Build systemsExplain systemsFollow curiosity
Selected work / five projects

Built to be
used.

Backend work, AI workflows, and the developer experience around them.

MeetStream / multilingual accessibility

Lingo

Turns multilingual meetings into live, speaker-attributed captions and searchable transcripts. Combines speech and translation providers with SQLite-backed jobs and a Chrome overlay.

Node.js · SQLite · Chrome extension · translation APIsExplore projectUp to eight transcript languages · VTT, SRT and portable exports
01 / 05
Meeting → meaning
Speaker 01

Good ideas deserve to be understood.

Illustrative caption · switch the language
MeetStream / agent infrastructure

Universal Agent Adapter

Unifies four agent frameworks behind one Python contract for MIA. The local pilot switches runtimes mid-meeting with streaming, cancellation, shared tools, and bounded conversation history.

Python · streaming · runtime adapters · MIAExplore projectOpenAI Agents SDK · Google ADK · CrewAI · LangGraph
02 / 05
One contract.
MIA / shared tools / streaming
Architecture study · select a runtime
open source · developer experience

MeetStream Labs

Turns meeting APIs into runnable Node.js examples across Zoom, Google Meet, and Teams, covering transcription, real-time media, agent workflows, and webhook reliability.

Webhooks · real-time media · Node.jsView sourceMeeting tools and developer resources
03 / 05
MeetStream Labs↗
01

Join a meeting

02

Receive a webhook

03

Stream the transcript

ZoomGoogle MeetTeams
Clone. Configure. Run.
Integration flow · visual study
meetstream · agent workflow

MIA Chat Agent

Deploys a hosted, wake-word-activated assistant into live meetings. Handles agent configuration, lifecycle webhooks, and confirmed bot removal, with answers delivered in meeting chat.

Webhooks · ngrok · Node.jsView sourceA meeting chat assistant
04 / 05
Hey assistant, what are the action items?
Hear the request.
Return the answer to the room.
Hosted agent workflow · visual study
personal build · ai product

Interview Copilot

A full-stack interview prototype with live captions, rule-based speech feedback, and session replay analytics. Handles caption revisions and deduplication, with transcripts persisted in MongoDB.

TypeScript · Socket.IO · MongoDB · MeetStream APIView sourceAn interview-feedback experiment
05 / 05
Interview / replay
Review the conversation

Notice the patterns.
Make the next answer clearer.

Session replay · visual study
The person behind the work

Make unfamiliar
work easier
to start.

I'm Arjun, a Computer Science student and AI Agent Automation Intern at MeetStream.ai.

Previously, I was a Developer Relations & Go-to-Market Intern at MeetStream.ai. My work now focuses on AI agent workflows and backend systems, with the same interest in helping developers move from an API reference to something running.

Start with something runnable.
Make the next step clear.
Let good defaults do some work.
Outside the build

Always another
rabbit hole.

Movies, music, puzzles, and the urge to figure out how the magic trick works.

Cinema aspect ratio
film

Frame by frame.

Villeneuve could film a parking lot and I'd cry about scale.

music

Sound opinions.

Ranking Kanye albums is basically a personality test at this point.

Knight puzzle. Move from b1 to g5 in three moves.
Reach g5 in three knight moves. Moves: 0.
puzzles

Happily stuck.

The pleasure of a hard problem, a wrong turn, and the very satisfying click when the pieces line up.

Field notes

Give the work
some context.

Writing demos, guides, and notes that make an integration easier to understand and use.

All writing