My Projects on Sumukh Acharya

Senior Vision AI

Fri, 15 Aug 2025 00:00:00 +0000

What is it?

Scan-n-Say is a mobile web app that helps elderly and visually impaired users shop independently at supermarkets. Point your phone camera at any product, and the app reads out a simple, friendly summary of what’s on the label — no squinting required.

Why I built it

Product labels are a nightmare — tiny fonts, dense ingredient lists, unfamiliar terms. For older adults or anyone with low vision, this makes supermarket shopping unnecessarily hard. I wanted to fix that with something genuinely useful: a tool that strips away the noise and just tells you what you need to know.

Grid Pulse

Tue, 01 Jul 2025 00:00:00 +0000

GitHub →

What is it?

Grid-Pulse is a real-time data pipeline that turns F1 race weekends into a live stock market. Every driver has a stock value that rises and falls based on their performance — qualifying laps, race positions, practice results — visualised on a live dashboard that updates as events unfold.

Why I built it

F1 is already obsessive about data. I wanted to explore what it would look like to model driver performance as a financial instrument — and use it as a real excuse to build a proper streaming data pipeline from scratch, not just a toy script.

URL Shortener

Sat, 15 Mar 2025 00:00:00 +0000

GitHub →

What is it?

A production-grade URL shortening service — the kind of system you’d actually deploy, not just demo. Built with Flask and Redis, containerised with Docker, and orchestrated on Kubernetes with autoscaling, load balancing, and health monitoring out of the box.

Why I built it

URL shorteners are a classic system design problem. I wanted to go beyond the basic implementation and actually build it the way it would run in production — with horizontal scaling, ingress routing, and the ability to handle real traffic spikes without falling over.

DFOS

Fri, 15 Nov 2024 00:00:00 +0000

GitHub →

What is it?

A multi-client file transfer system built from scratch in Python — no frameworks, just raw sockets. Clients authenticate, get their own isolated storage directory on the server, and can upload, download, preview, or delete files concurrently without stepping on each other.

Why I built it

I wanted to understand what actually happens under the hood when files move across a network — before abstractions like FTP clients or cloud SDKs hide all the interesting parts. Building it at the socket level made the concurrency and protocol design problems impossible to ignore.

Dynamic Fare Engine

Thu, 10 Oct 2024 00:00:00 +0000

GitHub →

What is it?

A fare forecasting system for RideWave, a fictional urban mobility company operating bikes, autos, and cars in Quahog City. Given three years of hourly fare data (2021–2023), I built predictive models for each vehicle type and combined them into an ensemble to power dynamic pricing recommendations.

Why I built it

Dynamic pricing is one of the more interesting applied ML problems — it sits at the intersection of time series forecasting, business strategy, and real operational constraints. I wanted to work through the full pipeline: exploratory analysis, model selection, feature engineering, and ensemble design, rather than just fitting a single off-the-shelf model.

Streamly

Tue, 10 Sep 2024 00:00:00 +0000

GitHub →

What is it?

A data analysis and modelling project for Streamly, a fictional streaming platform. Using their catalogue and engagement data, I analysed what content actually drives user retention and built regression models to predict it — giving the business a data-backed foundation for content decisions.

Why I built it

Streaming platforms live and die by retention. I was curious about what the data actually says when you cut through intuition — does budget correlate with ROI? Do certain genres retain users better than others? This project was about asking those questions rigorously and letting the analysis answer them.

Restaurant Ordering System

Mon, 15 Apr 2024 00:00:00 +0000

GitHub →

What is it?

A client-server restaurant ordering system built in Python using raw sockets, secured with SSL/TLS. Clients can browse the menu, place orders, and get a GST-calculated bill — all over an encrypted connection.

Why I built it

This was an early deep-dive into network programming. I wanted to understand how client-server communication actually works at the socket level, and adding SSL on top made it a practical exercise in securing that channel — not just making things talk, but making them talk safely.