Maaz Ullah

• Conducted Research and Development of AI-driven solutions for document digitization, data extraction, redaction,
and triage, using Python, LangChain, Gemini, and Google Vision OCR.
• Built specialized AI agents for data extraction and document digitization, integrating OCR, layout analysis, table
extraction, and object detection to enhance document parsing accuracy.
• Implemented RabbitMQ for efficient message queuing and API data transmission, enhancing system reliability
between backend and AI services.
• Architected asynchronous Gen AI API integrations to optimize system scalability and performance for large-scale
AI tasks.
• Deployed AI services using Docker Compose with multi-instance scaling for enhanced reliability and cloud
deployment.
• Created configurable evaluation tools with Streamlit for flexible AI model validation and performance
benchmarking.
• Applied prompt engineering and best practices in software development to ensure robust AI service
development.
• Designed and implemented Directed Acyclic Graphs (DAGs) to orchestrate AI agent pipelines and ensure
smooth workflow execution.
• Implemented and managed CI/CD pipelines for automated testing, integration, and deployment, ensuring faster
and more reliable software delivery.
• Collaborated with teams to enhance teamwork, streamline workflows, and drive project success.
• Automated key workflows, reducing manual effort and increasing productivity.

• Developed AI-driven solutions for resume parsing, comparison with job descriptions, and automated cover letter
generation using OpenAI, SpaCy, and Gensim.
• Created algorithms to match resumes with job descriptions, providing job fit scoring based on key qualifications
like skills, education, and experience.
• Built backend services using FastAPI and ReactJS, integrating real-time communication with WebSockets for
seamless job application processing.
• Deployed AI models and services on Google Cloud Platform (GCP) with Docker containerization, ensuring
scalability and reliability.
• Led cross-functional team collaboration across AI, QA, and Android developers, delivering the project on schedule.

• Developed an AI-powered real-time video analytics solution for retail management, focusing on features such
as people counting, face detection, gender classification, fall detection, heat map generation, and
queue management.
• Designed and implemented AI-driven video analytics using object detection algorithms like YoLo, RetinaNet,
Mask R-CNN, and DetectNet.
• Used DeepSort, StrongSort, and SiamMask for object tracking; ResNet and EfficientNet for object
classification.
• Developed a desktop application using CSharp and WPF forms, integrating OpenCV for real-time video
processing and data visualization.
• Implemented multi-threading and socket programming to optimize video processing performance and system
efficiency.
• Integrated video analytics with third-party applications using SDKs such as MileStone, Avigilon, and Qognify.
• Developed embedded applications for Axis Cameras using C++, WebSockets, and CMake, deploying via
Docker.

• Developed a web app for a Telco that focused on reporting and data visualization of users daily subscriptions using
Cassandra DB, Node.js (Express), and technologies like Bootstrap, D3.js, and Chart.js to create interactive
reports.
• Worked on the front-end of a video streaming web app for Telenors Goonj project using Swig (Template
Engine) and Node.js (Express).
• Applied web design and templating skills, along with graphic design work, converting PSD designs to HTML
templates.