A NOTE FROM MAVI

Hey, everyone.
I’m Mavi.

I’m Mavi / Anuj Giri. I like turning ideas into useful software and learning by building. Right now, my project work includes developing Waste2Worth and shaping DevReverse, alongside smaller practice builds and experiments. Come take a look around; each project tells you what it does, how it works, and what’s still in progress.

Hey! I’ll show you around.

INDEX / 01—08LIVE DOCUMENT
01Desktop toolsBUILD
02SustainabilityPRODUCT
03Interfaces & prototypesDESIGN
04Code practiceLEARNING

A small, static site. No frameworks, trackers, or external fonts.

PORTFOLIO / FIELD NOTESSCROLL TO EXPLORE

01 — SELECTED WORK

Projects, in context.

The story, system, and current state sit together. Open a project note for its architecture and limits.

01/projects/devreverse/WORKING ALPHA

DESKTOP TOOL · ELECTRON / JAVASCRIPT

A local-first browser and debugging workspace for inspecting a page, reviewing requests, and asking a local model about context you choose.

Make everyday browser inspection and local AI assistance available in one standalone desktop workspace.

Electron hosts the browser and inspection UI. Ollama runs separately on the same device; prompts use only the context selected for a question.

The private repository describes a working alpha: element and CSS inspection, session-only CSS previews, responsive viewports, opt-in console capture, request timing, local-model chat, and explicitly saved sessions. Source editing and performance analysis are future work.

Read the project note

Privacy by design: assistant prompts and selected context go to a loopback Ollama endpoint. Console capture starts off, sessions are saved only when requested, and there is no analytics or cloud AI endpoint.

Not yet included: source-code editing, performance analysis, advanced network controls, signed installers, and verified cross-platform releases. The repository is private, so there is no public installer or demo.

Open DevReverse README
02/projects/waste2worth/IN DEVELOPMENT

HACKATHON PRODUCT · WEB / AI

A sustainability platform that turns an item or material into reuse, repair, resale, donation, or upcycling next steps.

Help people find a practical next use for things they might otherwise throw away.

Static frontend pages call an Express backend. Configured AI providers analyze images or generate text ideas; Supabase handles authentication, persistence, and image storage.

The public repository contains frontend, backend, setup documentation, and an offline demo. A deployed service and live AI credentials are not verified here, so this is presented as an in-development codebase.

Read the project note

The repo documents image analysis, text-based upcycling ideas, multi-object scans, external inspiration-search links, marketplace and service pages, account pages, and an admin area. The provider chain is configurable; server credentials stay in backend environment variables.

Local setup and deployment steps are in the repository README. There is no verified live demo link.

Open Waste2Worth README
03/projects/kindred/OFFLINE PROTOTYPE

COMMUNITY DISCOVERY · HTML / CSS / JS

A single-page prototype for discovering community and NGO opportunities to donate, volunteer, or share useful items.

Bring community opportunities and lightweight guidance into one clear browsing flow.

A self-contained static page uses predefined sample opportunities, category filters, search, a details modal, and an offline question drawer.

The repository documents a browser-ready prototype that runs without a server or API key. Payment and personal-data flows are demonstrations only; nothing is submitted to a real service.

Read the project note

Kindred was created as a self-contained design prototype. It includes predefined opportunity cards, cause filters, progress and urgency labels, detail/support options, offline preset answers, and demo confirmation feedback.

It is not a live donation system and does not change InAmigos Foundation’s live website.

Open Kindred README
04/projects/c-programming/PRACTICE ARCHIVE

PROGRAMMING PRACTICE · C

Small programs covering arrays, pointers, structures, number manipulation, and basic file operations.

Build fluency by writing short programs around one language concept at a time.

Each exercise is an independent C source file that can be compiled with GCC or Clang.

The public repository documents the exercises and how to compile them. This is a learning collection rather than a single product.

Read the project note

The documented examples cover ASCII values, appending and creating files, array iteration, finding a maximum, structures, swapping values with pointers, and reversing a number.

Open C practice README
05/projects/c-integer-calculator/PRACTICE BUILD

SMALL PROGRAM · C

A first calculator program built around integer values, documented as an early learning exercise.

Turn basic control flow and arithmetic into a small interactive program.

A compact C program handles integer inputs and calculation operations.

The repository README describes integer-only behavior. Decimal support is explicitly future improvement, not a current feature.

Read the project note

The README says the calculator works with integer values and notes that improvements were planned as learning continued.

Open calculator README
06/projects/hackerrank-practice/PRACTICE ARCHIVE

CODING CHALLENGES · PRACTICE

A repository intended to collect solutions to HackerRank questions as they are completed.

Practice problem-solving through small, self-contained exercises.

Solutions are stored in a public GitHub repository; the repository description frames it as an ongoing collection.

No broader feature claims are made. The repo is a practice log.

Read the project note

The public README says the repository collects answers to questions solved on HackerRank. It is not presented as a finished application.

Open practice README
07/projects/private-ai-experiments/EXPERIMENTS

AI ASSISTANT NOTES · PYTHON

A set of assistant and local-model experiments captured across conversations, including chat memory and provider troubleshooting.

Explore an assistant that can chat and work with locally selected context.

Separate conversations cover an API-backed chat prototype and local Ollama / LlamaIndex experiments. They are not claimed as one completed product.

Conversation notes contain code and setup attempts, but no single verified repository, stable build, or demo link was found. Current functionality is unverified.

Read the project note

Conversation code references Groq and OpenAI-compatible chat calls, saved conversation history, and separate local-model / embedding / OCR installation experiments. This evidence does not establish a complete or currently working application.

Open experiment README
08/projects/offline-wallpaper/PLANNED

DESKTOP CONCEPT · LOW-POWER GRAPHICS

A design exploration for smooth, offline animated wallpapers with low background overhead.

Keep an animated desktop background responsive while limiting unnecessary work and network dependence.

The planning discussion favored operating-system graphics compositing and bounded frame memory rather than a custom driver.

This is a plan only. No source repository, implemented app, or benchmark has been verified, and smoothness on every device is not promised.

Read the project note

The conversation discussed a low-overhead, offline desktop wallpaper concept and clarified that RAM and GPU capability both matter. The plan is not a working build.

Open planning README

Status labels describe evidence in the linked repository or conversation notes. They do not imply independent runtime verification unless explicitly stated.

02 — APPROACH

Show the work.
Keep the edges honest.

This index borrows a case-study rhythm: start with the reason for a project, walk through how it is put together, and end with what is actually available today.

01

Builds
Describe the architecture and the functions documented in source.

02

Prototypes
Make demo-only behavior and missing live services visible.

03

Experiments
Keep plans and unverified work separate from shipped tools.

03 — CONTACT

Have a useful problem
to work on?

Reach out, browse the public work, or suggest what I should build next.

PROJECT / INTERACTIVE FIELD NOTE