Welcome

Manu R, full-stack engineer in Bangalore

Build. Break. Fix. Ship.

Turning caffeine into code for years. Full-stack apps, agentic workflows and cloud infra.

manu@maybecoded: ~

~ % whoami

Manu R, full-stack engineer based in Bangalore.

Turning caffeine into code for years. Full-stack apps, agentic workflows and cloud infra.

~ % cat focus.md

# Current focus

I turn your laptop, workstation or small server into a private AI workspace: local models, Hermes Agent, OpenClaw, tools, permissions, memory, and practical workflows for coding, research and automation.

runtime
local models, containers, GPU paths
agents
Hermes Agent, OpenClaw, tools, memory
delivery
apps, dashboards, infra, handoff

~ % ls ~/stack

  • react
  • flutter
  • typescript
  • python
  • fastapi
  • django
  • langgraph
  • ollama
  • gemini
  • three.js
  • firebase
  • supabase
  • postgres
  • mongodb
  • docker
  • terraform
  • gcp
  • cloudflare

~ % help

Try projects, open seein, contact or sudo hire manu.

Type a command such as help, projects or contact, then press Enter.

Projects

SeeIn

SeeIn

Turns a surgeon's one-line prompt into an interactive 3D operative scene, grounded in cited research and shown only after AI has inspected the actual render.

Research
Parallel research branches build a cited evidence dossier first.
Approve
Nothing is generated until the surgeon approves that dossier.
Verify
Every view is rendered in real Chromium and inspected by Gemini before it is shown.
  • TypeScript
  • Fastify
  • Three.js
  • Playwright
  • Gemini
  • ClickHouse
Case study Try it

Geodha

Geodha

Civic waste reporting for Bengaluru. Snap the problem, pin it on the map, and let AI draft the report.

Report
Photo, location and category in one short flow, with AI filling in the details.
Validate
Neighbours endorse, flag or resolve reports on a shared community map.
Escalate
Eligible reports go to BBMP as grievances through a linked Sahaaya account.
  • Expo
  • React Native
  • Firebase
  • Cloud Functions
  • OpenAI
  • Google Maps
Case study

Agentic health app

BLE sensor streamsLive readings
Meal contextWhat the reading means
LangGraph agentDecides what to recommend
Adaptive widgetsRecommendations in the app

Client work under NDA

Agentic health app

An agent workflow that reads live sensor data alongside meal context and turns it into recommendations that adapt.

Sense
Readings stream in from Bluetooth LE sensors in real time.
Context
Sensor data is paired with meal context before anything is suggested.
Reason
A LangGraph agent workflow decides what to recommend and when.
Adapt
Recommendation widgets in the app update as new data arrives.
  • LangGraph
  • Bluetooth LE
  • Agentic workflows

The product itself is private, but I'm happy to walk through the architecture.

CleanClik

CleanClik

Google ML-powered waste detection with QR-based bin categorization and a leaderboard loop.

Detect
Google ML Kit classifies waste live in the camera view, on device.
Dispose
Users are routed to the right bin, and a QR scan confirms the drop.
Compete
Points and a leaderboard keep people coming back.
  • Flutter
  • Supabase
  • Google ML Kit
Case study

SPARCD

SPARCD

Personalized goal app that breaks outcomes into adaptive micro-tasks for stronger follow-through.

Profile
An 8-dimension profile tunes task size, timing, tone and reinforcement.
Plan
Long-term goals become small daily actions tied to milestones.
Adapt
When momentum drops, tasks get simpler and timing shifts.
  • Flutter
  • FastAPI
  • MongoDB
  • Redis
  • Firebase
  • RevenueCat
Case study

NexxLynx

NexxLynx

Matrix client with a local LLM runtime and a Postmoogle email bridge.

Chat
Matrix-native secure messaging with voice and video.
Email
A Postmoogle bridge brings email threads into rooms, no forwarding.
Local AI
A local runtime summarizes threads and extracts action items privately.
  • Flutter
  • Synapse
  • PostgreSQL
  • Azure
  • Cactus AI
Case study

Services

Local AI assistants

Hermes Agent, OpenClaw and local models wired up on your own machine.

I install the runtime, connect models, configure tools and permissions, add memory where it helps, and make the assistant usable for coding, research, automation and day-to-day operations.

  • Hermes Agent
  • OpenClaw
  • Ollama
  • llama.cpp
  • Docker
  • MCP

Full-stack apps

Web and mobile apps built for speed, scale and maintainability.

I design and build complete systems from UX flows to APIs, auth, payments, data modeling and deployment, so your team can ship faster and iterate safely.

  • React
  • Flutter
  • Django
  • FastAPI
  • GraphQL
  • MongoDB
  • LangGraph
  • vLLM
  • Agentic apps

Cloud and infra

Cloud and local infrastructure for apps, agents, workers and model workloads.

I set up CI/CD, infrastructure as code, monitoring, alerting, GPU-aware deployment paths and rollback-ready releases, so systems stay predictable.

  • AWS
  • Google Cloud
  • Azure
  • Terraform
  • CI/CD
  • Local infra

Tech consulting

Got an idea and want to vibe code it? Let's get the architecture right first.

We map the scope, stack, data model, APIs, auth, deployment path and risks before the build starts. Then you can vibe code with a clearer path and avoid a lot of future cleanup.

  • Product scope
  • Architecture
  • Stack choice
  • Data model
  • API design
  • Auth flow
  • Deploy plan

Debug and fix

For stuck builds, vibe-coded prototypes and products that need to get fixed and launched.

I trace broken flows, clean up rough generated code, fix frontend, backend and data issues, stabilize deploys, and get the project into a shippable state.

  • Bug triage
  • Code cleanup
  • React
  • API fixes
  • DB fixes
  • Deploys
  • Performance
  • Launch

New message

To manu@maybecoded.com

Send the build, bug, repo, machine or cloud limits, and what should be working first. I can help you scope it, fix it, or turn it into a shippable system.

Add a line about what you need before sending.

  • @manudotdev
  • manuu-r
  • Signal: manu_r.93
  • GitHub
  • Writing on Medium