About The Company

Megafuse is a INR 100 Crore+ revenue semi-conductor distributor that acts as a platform between sellers of excess semi-conductor inventory and buyers needing the same inventory at a low cost. Megafuse is currently headed by Shrenik Jain, a dynamic CEO who passed out of IIT Bombay and worked at the esteemed BCG as a management consultant. Under Shrenik, Megafuse has consistently served 300+ clients, offloaded inventory from 50+ component manufacturers, all in a period of 20 years.

‍

Problem Statement

Buyers of Semi-Conductor equipment often pay 50 to 100% more for parts since they aren't aware of more affordable products offering the same specifications.

Additionally, they also need to face turnaround times of more than 2 days in placing orders and a few weeks in receiving them.

‍

Approach

We decided to build a tool for Megafuse that is able to recommend affordable parts within the same specifications as the ones that their customers are currently using.

For this, we had to create our own database of parts, to the tune of 10M+ parts, each with 20+ specifications.

Next, we had to create a gold dataset of recommended parts taking help from domain experts, and train our own model on it. Basis the data, we aimed to develop an AI offering 99%+ accuracy of recommendation.

‍

Key Challenges

Creating a database of 10M+ components included scraping 3 online databases each with 7M+ items, cleaning and merging the data, and then creating fields according to the customer's definitions. All 3 databases had differently structured data, and it required significant consultation from Subject Matter Experts to normalize the data in a standard format.

The number of parameters for each category of products was in the order of 100, and the ML algorithm was not explainable. To make it explainable meant trading off on the quality, and so we did that.

‍