Asif Hasan

A payment processing web app for patients of hospitals and clinics.

I was a senior software engineer and worked on the front end, back end, and development of their CI/CD pipeline and Kubernetes configurations.

THE PROBLEM
For a payment processing web app, different hospitals and clinics needed to customize what information they wanted to collect and track from their patients. The app also needed to support the configuration of derived fields and how those should be calculated. For example, the total amount of a line item can be calculated from other data fields in that line item, e.g. unit price, quantity, discount code, etc. Previously, the hospital clients needed to contact the development team.

MY SOLUTION
• I delivered a secured and powerful equation editor for the derived fields. One could use other fields from the line items as variables. In addition, I implemented support for functions with which they could store an equation in the library for later use inside another equation.
• I provided a set of commonly used equations as functions in the default library.
• I provided a function that could use data from the back end, which was very useful for calculating discount amounts from a customizable set of discount codes and rules.

PROBLEM
The API back end was failing to serve the front end as the number of users grew. It was developed on a monolithic architecture and was throwing errors due to the memory getting filled or was serving the requests with very high latency due to starvation of threads. I measured the server to be able to serve only 50 transactions per second (TPS), whereas the need was to serve over 500 TPS.

SOLUTION
I completely re-designed the back end to be horizontally scalable. I implemented an auto-scaling mechanism that scaled down the infrastructure to serve only 50 TPS on low traffic but could serve 2K TPS on peak demand. We verified all these numbers using load testing simulating actual traffic distribution.

PROCESS
I used AWS services and broke the monolithic architecture into multiple microservices. The microservices were deployed to AWS Lambda and AWS ECS clusters. I used auto-scaling and the site’s traffic pattern for the ECS cluster to design auto-scaling steps. For the database, I used DynamoDB, and for the storage, I used AWS S3, both of which can serve demands at scale. I made a lot of the back-end processes asynchronous and event-driven using the Pub/Sub mechanism with AWS SQS and SNS and DynamoDB streams.

I developed a fully functional SIP dialer that can register to any SIP server. The app had the following major features:
• Make/Receive calls
• DTMF support to interact with IVR
• Messaging
• Conference support
• Integration with native notification system for incoming calls
etc.
I used Java and Linphone-Android.

I developed the full stack of a virtual phone system for Android and iOS that includes the mobile dialer and the server infrastructure.

For the front end, I used React Native, TypeScript, and Twilio SDK.
For the back-end, I used a Twilio server for VoIP and AWS, Express.js, and TypeScript for API service infrastructure.

The phone app supports the following major features:
• Buy new phone numbers
• User Auth with OAuth (email/pass, social: Google, Facebook)
• Multiple phone numbers
• Make/Receive calls
• Send/Receive SMS/MMS
• Group Chat
• Integration with native call notification and history
• Sync native contacts
• Call blocking functionality

• Scraped website data from 20,000 hotel websites across the US.
• Used the Playwright framework with TypeScript.
• Scraped various source data formats, including HTML, PDF documents, etc.
• Developed a continuous monitoring script to check for source data changes.
• Deployed in AWS.
• Exposed the scraped data through API.

• Developed a web scraping project to scrape post data from a popular ad listing site.
• Used the Playwright framework with TypeScript.
• Deployed the continuous running scraping project in AWS with APIs to expose the scraped data.
• Developed a detailed reporting dashboard to monitor the success/failure of the scraped sources.
• Used a deep learning-powered watermark removal tool to remove watermarks from images.

Developed a robust and fault-tolerant workflow to populate data of the hotels across the USA. I built the UI components and corresponding workflows for the feature, developed data custom APIs, and integrated existing 3rd-party APIs such as Google Places API.

Developed the UI components to show a media gallery from the user's Instagram posts and the corresponding workflows and API calls. I also built the user authentication and authorization solution for this project on the Bubble.io platform and added support for social auth, e.g., Facebook and Instagram, using OAuth 2.0.

Developed a React Native mobile app with an LLM-powered Azure back end for group task management.

Tech Stack:
• Front end: React Native, TypeScript
• Back end: C#, .NET Core, OpenAI, Claude
• Cloud: Azure (Functions, Database for PostgreSQL, API Management, Load Balancer, Queue Storage)
• IaC: Terraform
• CI/CD: Jenkins, Ansible

Key Features:
1. Real-time task synchronization
2. AI-powered task prioritization
3. Intelligent notifications

Azure Implementation:
• Azure Functions for a serverless back end
• Azure Database for PostgreSQL for data storage
• Azure API Management for API security
• Azure Load Balancer for high availability
• Azure Queue Storage for async processing
• Terraform for infrastructure as code

I implemented a CI/CD pipeline with Jenkins and Ansible, automating Azure deployments. Infrastructure is managed via Terraform.

This project demonstrates expertise in full-stack development, cloud architecture, IaC, and AI integration, aligning with the warehouse management and robotics project needs. It showcases proficiency in Azure services, particularly Azure Functions, and highlights experience with modern development practices and tools.

Developed a sophisticated multi-agent chat system using Python and AutoGen, integrating various AI technologies to enhance communication efficiency and user experience.

Key Achievements:
• Implemented multi-agent architecture, reducing response time by 30%
• Utilized Claude for text-based replies and prompt generation, improving efficiency by 25%
• Enhanced synchronization between the front and back end, boosting message flow by 40%
• Integrated corpus in system prompts, increasing interaction relevance by 35%
• Optimized image analysis/generation using Fireworks, improving speed by 50%
• Implemented streaming capabilities, decreasing turnaround time by 20%
• Employed prompt caching, enhancing response times by 65%
• Improved UX with front-end enhancements and Notetaker feature
• Incorporated Together for tag analysis, enhancing conversation quality

Technologies:
Python, AutoGen, Claude AI, Fireworks, Together AI, React, Azure, AWS

This project demonstrates expertise in AI integration, multi-agent systems, and full-stack development. It showcases proficiency in optimizing AI-driven applications for performance and user experience, aligning with requirements for complex system development in warehouse management and robotics.