Archit Mathur

A cross-platform mobile app for sellers to receive critical alerts about their Amazon store, such as SKU delisting, poor customer reviews, and low inventory count. It allows sellers to resolve things quickly and avoid adverse after-effects like account health drops, store closure, fee increases, etc. Users start receiving notifications when they subscribe to a plan and link their Amazon store with the app.

I built this product end-to-end using Node.js, TypeScript, NestJS, and MongoDB for the back end hosted on ECS/Fargate; TypeScript, React, and Next.js for the website hosted on Vercel; and TypeScript, React Native, and React Native Elements for the mobile app available on the major app stores. Subscriptions are powered by the Google Play billing system for Android and Apple IAP for iOS, and notifications are sent to users using Firebase.

A web platform written end-to-end in TypeScript/Node.js, which allowed UK residents to compare their current energy or broadband provider with other providers in the market. Users could also initiate a switch to one of those providers.

I was the primary developer in the team and worked on the following:
• Developed the front end using Next.js and Material UI.
• Established a monorepo with TurboRepo, encompassing both platform front ends and other UI and core libraries.
• Engineered the back end with NestJS.
• Implemented CI/CD pipelines on GitHub for front-end deployment on Vercel and back-end deployment on Amazon ECS.
• Orchestrated the setup of production and staging environments on AWS, leveraging AWS CDK.
• Configured OAuth with Auth0 for user authentication.
• Managed DNS configuration across GoDaddy, Amazon SES, AWS ALB, and Vercel.
• Integrated Google Analytics with custom events.

The platform allowed token holders of a specific token issued by the client to move them from the Waves blockchain network to the Polygon blockchain network.

I was the back-end engineer and liaised with the front-end engineer, product manager, solutions architect, and tech lead to build out the server-side architecture from scratch. This involved:
• Developing RESTful APIs with NestJS.
• Implementing cron jobs to synchronize market data sourced from CoinGecko, orchestrated through Kubernetes.
• Establishing Kafka consumers to ingest block updates seamlessly.
• Integrating with various 3rd-party APIs like Auth0, Waves REST API, Polygon JSON-RPCs, CoinGecko REST API, ConsenSys Orchestrate, and more.

A platform to manage warehouse inventory and sync it across multiple Amazon stores. I was the sole engineer and built it out with TypeScript/Next.js on the front end and Go on the back end. The solution had features like item creation/deletion/updates/syncing, along with other quality of life stuff like critical issue alerts (around SKU delisting and the like) for all the connected stores in one place.

My contributions encompassed:

• Full-stack development of the front end using Next.js and Material UI.
• Development of the entire back end infrastructure using Go, with DynamoDB as the database.
• Integration of Amazon Selling Partner API (SP-API) and ShipHero GraphQL API for seamless data synchronization.
• Containerization of the entire application and establishment of the deployment pipeline for streamlined deployment processes.

This pipeline was meant for smallcase's communications infrastructure which received bursts of requests from clients and integrated with flaky B2B 3rd-party APIs. This often led to timeouts, 429s, and a general degradation of quality of service. The idea with this throttling pipeline was to ingest requests at any rate from clients and invoke 3rd-party APIs at a constant and predefined rate.

My role was to build this pipeline end-to-end in a Node.js service. I designed a custom scheduling algorithm powered by Kafka, which already stored the client requests, and Redis, to manage state across the service cluster.
The algorithm equispaced 3rd-party API invocations by using backoff on client requests consumed from Kafka. Redis stored the last time an API was invoked and the time to backoff a client request was calculated using it and a predefined gap (like 2s). As soon as all the requests were processed serially for a Kafka batch, their offsets were committed.

Introduction of this pipeline reduced internal systems' load metrics by around 98% and SLA-based errors from 3rd-party APIs by around 70%.

FabBits is a standalone cross-platform and open-source software capable of finding certain interesting bits from movies, tv shows, and sports. It is able to detect the following:

• Action sequences in movies and shows
• Summary of movies and shows
• Actor-specific scenes in movies and shows
• Jokes in sitcoms
• Goals in soccer
• Three-pointers in basketball

This project was a result of my time with CCExtractor during Google Summer of Code 2018 and was accompanied by a series of blog posts.

AWS CDK is used to create the AWS infrastructure required for third parties to upload images (JPEGs) to a private S3 bucket. This stack also sends an email every time an image is uploaded to the bucket.

This AWS Lambda ingests Google Calendar event push notifications and adds them to a Google Sheets spreadsheet post-filtering for the useful ones. It's written in Go and uses an AWS Lambda Function URL to receive the webhooks.

Tweeder weeds out tweets of a certain kind by specific authors that you don't want to see on your timeline. This serves as a great tool, for example, to filter out the tweets by a domain expert that you find irrelevant or not of your liking but still see the tweets on the topic for which you followed them.

A REST API that allows users to register and manage their soccer team. Users can trade players with each other, add uncapped players to their team, and cancel contracts of players from their team—full documentation in the accompanying Postman collection.

An accompanying, WIP, SPA is this - https://github.com/achie27/fantasy-soccer-spa

I contributed to Mattermost's back-end services that power their chat app, Focalboard, and the GitHub plugin.

This involved writing and updating the business logic of APIs, interacting with the database, specifically Postgres or MySQL, and removing tech debt for performance or ease of maintenance gains. All of this was done in Go.

blobCloud is a React single-page app that can save public web files directly to a user's Google Drive. After the file saves, it notifies the user via email. It also maintains the downloaded and saved history.