Tayyab is a highly skilled software engineer and accomplished researcher with a proven track record in the computer software industry. He holds a master's degree in computer science and possesses over six years of invaluable experience in diverse technology stacks, successfully delivering dynamic products across varied market sectors, particularly in the healthcare domain. Tayyab's expertise lies in data science and he excels in full-stack web development, with a primary focus on leveraging Python, Microsoft technologies, React, and Angular.
AI, ML & LLM
Pytorch
Machine Learning
Deep Learning
Neural Networks
Convolutional Neural Networks (CNN)
GPT
Generative Pre-trained Transformers (GPT)
Unity (IoC Container)
OpenAI GPT-4 API
OpenAI GPT-3 API
BERT
SBERT
Email APIs
Large Language Models (LLMs)
OpenAI
Together.ai
Generative AI
Worked as a full-stack developer, creating features and fixing the issues with the existing version of the application.
Architected the new back-end architecture, improving the existing codebase.
Developed multiple modules, including token-based authorization and syncing products from Gmail in real time using Gmail API.
Created a large language mode (LLM)-based email parser as a fallback strategy for extracting clothing article details from emails if the existing rules-based system cannot parse the email.
Developed a complete ML pipeline from processing and transcribing audio calls to speaker diarization, and generating a score based on configurable domain-related questions for scoring a call between a client and a customer service person using LLMs.
Crafted a FastAPI-based product around the ML solution.
Architected the application, adding authentication and multi-tenancy support and creating endpoints for exposing data for analytics.
Created a module to provide a queue-based mechanism for processing call-analysis requests.
Build integration with Stripe for managing user payments.
Worked as a full-stack developer, primarily on the admin dashboard application. Developed optimized (porting from VB.NET) back-end APIs to expose data using ASP.NET Core 6.
Revamped the entire front-end application in Angular, working on the authentication screens, admin dashboard, and admin form controls for adding/updating/deleting members, counselors, and applicants.
Worked on the analytics dashboard for admin using Google charts to show stats, including graphs to indicate the number of enrolled applicants in the country and the US, counts of applicants with counselors, and applications of member institutions.
Worked on the AI API, creating a layer over GPT to act as a personality specified, not only popular but commoners as well. Worked on several improvement modules to produce custom responses to several edge cases where no response from GPT is required.
Supervised the development of the application's back and front end using FastAPI and React, designing the architecture and managing the development tasks.
I represented my company in different AI meetups and different panel talks and trainings at different Universities.
Started and managed the data science team, training fresh resources and helping them develop research and development skills in machine learning, data engineering, and data analysis. Initiated many research and development projects by analyzing the company's existing products and their data.
Won the best team of the year award for the project related to automated data migrations from super bill forms. Led the development of many end-to-end extensions to the company's electronic health record software, adding assistive features that helped improve the overall practitioners' user experience and decision process.
Created a training outline for hiring new Python development and machine learning talent.
Collaborated with many accomplished academic researchers on research publications about applying artificial intelligence and machine learning in healthcare.
Engaged in the full-stack development of NOVATRAQ and FUNDINGTRAQ for a US-based client. Received the client's appreciation award multiple times, including one during the COVID-19 pandemic when our team delivered the Paycheck Protection Program (PPP) loan integration within a month.
Built the full-stack of a Dubai healthcare IoT, which helped automate the vitals recording process and reduce the nursing staff's workload.
Developed the extension of NextHRM, a company-owned product.
Created a management module to view, add, update, and remove rooms and workstations available on office premises.
Developed the back-end of Vicenna HealthCloud, a cloud-based hospital management system, primarily working with ASP.NET Web API 2 and Entity Framework.
Researched and developed the application's proofs of concept to comply with Health Level Seven (HL7) international standards, including using Mirth Connect as middleware to convert incoming data of different HL7 versions.
Executed various integrations, including services for communication between HealthCloud and the Dynamics 365 ERP system. Standardized communication between HealthCloud and the laboratory system.
Won the Employee of the Month award for my work on the application's optimizations and standardization.
Developed a cloud-based Hospital Management System (HMS) offering Software as a Service (SaaS) with key features including high configurability, cloud-based computational processes, and a single-page web application user interface.
Designed web services for data exposure to the frontend and ensured application's compliance with HL7 standards. Also, integrated other established systems like Dynamics 365 for seamless synchronization of patient details.
Enhanced application code to optimize its performance during high usage. Introduced microservices to extract specific functionalities and explored the implementation of Cosmos DB for efficient data retrieval, surpassing traditional relational databases.
Developed a software solution for seamless integration, automating patient vital readings for efficient nurse workloads with a management portal and a reading room application.
Utilized various technologies such as ASP.NET MVC, Dapper, HTML, CSS, Bootstrap, JavaScript, C#, jQuery, WPF, SQLite, and Material Design WPF in the creation of the management portal and the reading room application.
Implemented features like Opentok WebRTC-powered nurse auto-ride for audio and video communication, Windows text-to-speech API for patient instructions, and an HL7-compliant REST API server for widespread healthcare system compatibility.
Experienced full-stack developer specialising in the integration module for FICO LiquidCredit Small Business Scoring Service, with a focus on designing and implementing components for processing credit scores in loan applications.
Developed credit memo customizations for specific lenders using Kendo jQuery UI components and enhanced existing UI controls. Utilized Telerik Reporting for creating system documents and reports modules related to PPP loan forgiveness summaries and credit recommendations.
Contributed to the development of a Telerik RAD Controls-based PDF engine, ensuring smooth mapping of SQL table data to and from the required fields of PDF forms.
Developed the backend of the biometric attendance management portal, creating Windows services that operate on individual attendance terminals, recording and synchronizing employees' time in and out, with thumbprint verification.
Worked as a full-stack developer to construct a meeting room and workstation management module using ASP.NET MVC, enabling users to manage rooms and workstations within the office premises.
Designed and implemented the meeting room and workstation assignment module, addressing use cases such as booking meeting rooms, managing attendees, authorizing meetings, notifying schedule changes, managing workstations, and notifying workstation assignments.
Developer designed and implemented a system for processing services like mobile phone top-ups and bill payments. This was achieved by developing a backend server for a self-service terminal featuring REST APIs.
Developer integrated multiple hardware devices such as a payout device, cash validator, card reader, and printer into the system. They also created an Android application for agent verification, terminal maintenance management, and a WPF application for better user interaction.
Developer ensured uninterrupted functionality of the system even without network connectivity through a transaction sync manager and further provided a web portal for managing administrative actions.
Developed a cloud-based application and subsystem of NOVATRAQ to facilitate necessary activities during the SBA loan process including customizable loan application wizards, application monitoring, and document management.
Designed loan application interview wizards for SBA, SBA PPP, and SBA PPP forgiveness using Kendo jQuery UI, Backbone.js, and Marionette, and applied ASP.NET Web API with Entity Framework for endpoint APIs.
Implemented client-side management system with SignalR and developed product match, self-registration modules, and documents template module using Rackspace.net for SBA PPP loans and associated forgiveness applications.
The application is purpose-built for Pakistani products and gives users suggestions for alternative medications along with an easy finder for nearby pharmacies that have desired medicines. Also includes a dedicated interface for pharmaceutical businesses for efficient management.
Being a highly skilled full-stack developer, separate ASP.NET MVC and Web API servers were developed, and secure communication ensured through custom, claim-based authorization and token-based authentication. Leveraged Google Maps API for search functionality and the pharmacy locator module.
Expanded the functionality of the application through utilizing Selenium for .NET and other data cleaning methods for creating a large dataset of alternative medicines specific to Pakistani products. Successfully designed and implemented manufacturer and pharmacy management modules using ASP.NET MVC.
I led the development of a Deep Learning-powered image classifier which can accurately identify over 200 distinct bird species using Convolutional Neural Networks (CNNs).
I fine-tuned models and applied K-fold cross-validation techniques for model refinement. The most effective models used the Inceptionv3 architecture training on the ImageNet dataset.
I amalgamated multiple pre-existing labelled datasets, including the CUB and NABird datasets to augment my own dataset.
Spearheaded the development of an AI-driven resume parsing system for efficient candidate profiling and suitability scoring based on job descriptions
Lead the creation and architecture of the Parsing API and other modules utilising Python, test-driven development, compliance with PEP standards and implementation of Flake8
Directed the R&D team in preparing training data for fine-tuning the Spacy NER (BERT) model and construction of an ML model leveraging style occurrences for resume classification
Developed an advanced deep learning system for precise face detection and recognition, enabling seamless tracking of patients in a medical facility, and accurately timing therapy sessions.
Utilized ArcFace and Facenet512 for constructing robust face recognition and SSD for dynamic face detection in real-time video; collaborated in designing application architecture, including a streamlined patient registration process using face recognition and an edge-based system for improved communication with existing EHR system.
Managed the development of a FastAPI-based monolithic server to handle various endpoints for recognition purposes.
Expert in creating high-quality content and crafting compelling profiles. Worked closely with machine learning engineers to develop sophisticated bots that embody the qualities of renowned personalities. These bots can handle communication, product endorsement, and marketing activities.
Conducted extensive research on diverse embedding models and optimized the process by implementing solutions like Milvus for vector searching and open-source options like Llama-V2-7B-chat. Developed a backend system based on FastAPI and architected a multi-tenancy infrastructure for accommodating multiple personalities.
Integrated text-to-speech and speech-to-text modules using Whisper and ElevenLabs respectively, adding audio input and output functionalities to the offering.
Developed an ingredient extraction system to retrieve data from restaurant menus, connecting retailers and restaurants by identifying ingredients sold and consumed.
Constructed a data preprocessing pipeline for cleansing and normalizing the information and devised a two-step ingredients extractor, using the Language Learning Model (LLM) to fill in missing information.
The project is active and evolving, aiming to create an in-house machine learning model to replace LLMs and provide a more scalable solution.
Developed a machine learning system to identify duplicate products across various food delivery platforms enabling price comparison and detection of missing products.
Served as the lead data scientist and designed a system to extract product attributes, creating a semi-supervised tagged dataset used for training machine learning models.
The system displayed a 98% accuracy rate and processed over 3 million product records, facilitating the identification of common stores across platforms and elimination of duplicate store entries.
Developed an add-on for Reki platform that provides personalized movie recommendations using LLMs to solve cold start problems
Constructed an end-to-end solution with OpenAI GPT-4 for comprehensive movie information extraction, also experimented with Mistral and Llama models to explore cost-effective alternatives
Implemented a secure API key-based authentication system to ensure secure communication between Reki app and the recommender
Developed a machine learning and natural language processing solution to extract retail product data for sales analysis and market assessment from platforms like UberEATS, Zomato, Hunger-Station, and Jahez.
Constructed an advanced ETL pipeline using cutting-edge machine learning techniques to identify and extract key attributes. Improved system performance significantly by replacing pandas with Dask data frames.
Implemented a streamlined data cleansing and normalization pipeline with NLP methods for unlabeled data and created dedicated modules for confidence calculation, post-processing, and verification. Successfully tagged over 10 million records with the final system.
The developer led the creation of a machine learning system to identify duplicate stores from various food delivery platforms, forming a foundation for data analysis and correction.
The developer designed a Natural Language Processing-based system, tagged 40,000 examples with human help, experimented with various machine learning and deep learning models, but Siamese networks did not meet the desired objectives.
Geo-fencing techniques and geo-hashing libraries were implemented for efficient grouping of stores, which enhanced system efficiency and eliminated duplicate store entries. The system processed over 500,000 records with a 96% accuracy rate.
Developed an AI tool to analyze and score calls between customer service agents and clients to track customer satisfaction and manage call quality.
Designed and implemented a machine-learning pipeline for processing and transcribing audio calls, speaker identification, and generating domain-specific call scores.
Incorporated multi-tenant authentication, data analytics endpoints, queue-based call analysis processing, and Stripe integration for user payments in the FastAPI application.
Provided a platform for managing a digital closet with features like purchase tracking, shopping insights analysis, and closet listings organization.
Worked as a full-stack developer; created new features, fixed issues in the existing application, and improved backend architecture.
Developed modules for token-based authorization and real-time syncing of products from Gmail API. Also created LLM-based email parser as a fallback strategy for extracting clothing article details from emails.
