An entrepreneur at heart, Abhishek is experienced in developing scalable and maintainable web applications. He likes to dive deep into a problem, investigate it thoroughly, and then come up with the most appropriate solution. His master's degree in mathematics and computing also comes in handy when solving the most complex problems.
Developed a fully configurable platform from scratch to build and deploy applications for any business use case. It included configurable modules for auth, access control, schema enforcement, business rules, and CRUD on the Neo4j database.
Worked on a new platform that reduced certain complex queries' time from minutes on the client's existing platform to sub-seconds on the new one.
Built a demo front end for the platform to be used for a sales showcase.
Developed a rule engine that is used to calculate the amount of delivery charge that is applied in order. The key features of the rule engine are a validation of created rule and fast execution for the given input.
Led a team of two developers to build a dynamic delivery area system. The system decides, at runtime, the areas where a particular merchant can deliver food taking into account factors such as the gap between consumer demand and delivery-boy supply.
Led the development of the serviceability engine which is responsible for calculating whether or not a particular merchant can be shown available to a specific user. The engine can handle a peak throughput of 600k RPM.
Implemented the time-based one-time password algorithm (TOTP) which allowed validation without using a database and for higher reliability.
Implemented a novel missed-call-based OTP delivery mechanism.
Optimized the real-time routing of requests to multiple SMS gateways based on their performance and users’ geolocation. Used an open Geo-IP database to get the user's location from a request IP address.
Built a secure URL shortening service (CMPR.es) with a safeguard from malicious crawlers.
Constructed a JavaScript library to allow the front-end integration with just four lines of code.
Developed and scaled a Facebook-like social feed system.
Used best practices to build robust and scalable services for CRUD, social actions, and permissions.
Built an ad system to deliver sponsored content based on probabilities generated by bid scores.
Created an ML-based bill image analysis system to extract metadata from uploaded bills.
Implemented a "rolling window Levenshtein" string similarity approach to tackle garbage text read by the OCR.
Implemented a Graph-based song similarity algorithm to generate real-time playlists using user feedback.
Exploited YouTube’s public playlists to solve the cold start problem.
Developed plug-n-play hardware to stream music in automobiles. Designed a stream protocol to interface between Raspberry Pi and ESP8266 to handle speech and music player controls.
Built a speech recognition server to convert a user’s speech to text and aid in song searches.
Developed the back end and algorithms of the electric vehicle planning and assessment tool.
Built the back end using Java 8 on Google AppEngine (for CRUD).
Created a data extraction layer, to read structured data from fleet operator schedules (.xlsx files); Web Worker.
Simulated bus schedules to check for EV viability; Web Worker.
Wrote an algorithm to minimize the required charging locations to run the electric fleet; Web Worker.
Developed a generic integration and regression testing framework for Amazon.
Wrote a parser to create a test API and service calls using production logs (legacy format).
Received a pre-placement offer to join as a full-time engineer.
A Flutter-based app that allows users to compare prices of multiple ride-hailing services in one go and then proceed to book via an in-app browser. Ride prices are fetched from the web apps of service providers by adding hooks to Ajax calls sent to their respective servers.
A Flutter-based cross-platform app (Android and iOS) that aims to simplify the online shopping experience on mobile devices. It solves two critical pain points for a shopper:• Shortlisting and comparing products between multiple platforms.• Discussing with friends to get their opinion.Developed an eCommerce scraping system using Node.js and Puppeteer. The back end was deployed on preemptible instances of GCP to leverage changing IP addresses.
DrawGuess is an online, cross-platform port of the classic drawing and guessing game—Telephone Pictionary. It also features a WebRTC-based group voice call for players to interact together.The game has garnered over 75,000 user sketches within three months of its launch.
A fast and scalable URL shortener service built using AWS Lambda and DynamoDB. It supports setting a TTL on shortened URLs and can redirect to different URLs (e.g., for Android and iOS) using user-agent-based detection.
A Flutter-based app to simplify address sharing for users. A user can save their address along with the GPS coordinates in this app. Saved addresses can be shared with people over WhatsApp, SMS, email, etc. A public web page is generated for each shared address. This page has options for a user to see directions or book a cab with one click.
IdentifyMe aims to solve Identity-as-a-service.To start with, it offers an end-to-end phone verification service that allows organizations to integrate login and signup via phone as easily and reliably as a social login via Google or Facebook. It also solves the SMS reliability issue by using a novel, missed-call-based OTP delivery mechanism, which is inherently more reliable and faster.
Musicmonk is an intelligent music-streaming service that automatically plays music to your taste.It is a departure from playlist-based music toward mood-based music. Users start by selecting a song from a pre-computed list of distinct songs based on the user's history. As the session progresses, the user is presented with music based on real-time feedback (duration and skip).A new interface, still under development, can be found on the website.
