Saife Shuhaib Md.

I am a dedicated software engineer with a strong focus on AI, Machine Learning, and Backend Development. I have experience building intelligent systems that solve real-world problems, from creating AI-powered solutions like Generative AI and NLP to developing reliable and scalable backend services. With a solid foundation in Python and JavaScript, I work with tools and frameworks such as Django, FastAPI and PyTorch to deliver impactful and efficient results. I enjoy leveraging technology to simplify complex challenges and create seamless user experiences.

Key Contributions

• Designed the system for processing text data and storing it in a vector database.
• Developed a module to retrieve state-specific legal data from the vector database.
• Designed a logging system for the scraper used to extract legal data.
• Optimized vector query accuracy through data cleaning and query augmentation.
• Improved citation module accuracy by 60%.
• Implemented multithreading for parallel execution of independent API calls within the system.
• Reduced overall system response time from 1.5 minutes to 30 seconds, a 67% reduction.

• Python
• Django
• PostgreSQL
• AWS (RDS, EC2)
• Supabase
• FAISS
• Docker

Key Contributions

• Implemented meeting transcription versioning and tagging systems.
• Designed APIs for generating various file notes from user transcripts.
• Migrated all database tables and data from the dev environment to production using pg_dump.
• Helped transition from session-based user settings to fetching them from the database.
• Increased user satisfaction by 80% and eliminated manual data entry at each login.
• Created APIs for letting admin modify system prompts from the front-end.
• Increased admin productivity and completely eliminated developer dependency for prompt modifications (100% reduction).

• JavaScript
• Express
• Knex.js
• PostgreSQL
• AWS (RDS, EC2)
• Docker
• Assembly AI

Key Contributions

• Developed a module to detect comic book panels and speech/thought bubbles.
• Trained an object detection model to identify object areas and extract polygon coordinates.
• Conducted project time estimation, resource planning, and resource allocation.
• Implemented preprocessing techniques for comic book images to enhance detection accuracy.
• Optimized model performance for inference and deployment.
• Reduced translation error by 25%.

• Python
• PyTorch
• YOLOv8
• OpenCV
• ONNX Runtime
• Roboflow

Key Contributions

• Created model loaders and converted models to OpenVINO format for edge device inference.
• Packaged emotion and landmark detection models into DLL files for seamless integration with desktop apps.
• Applied multithreading techniques to improve processing efficiency and speed.
• Implemented frame-skipping to reduce computational load while maintaining accuracy.
• Ensured the solution provided real-time analysis within the constraints of edge device hardware.
• Improved performance by 40% through multithreading and frame skipping.

• C++
• OpenVINO
• OpenCV

Key Contributions

• Developed a system to securely store face embeddings.
• Implemented homomorphic encryption using Microsoft SEAL to perform secure recognition on encrypted data.
• Integrated the encryption module with an Android app using Java Native Interface.
• Optimized encryption settings to minimize file size and decryption time.
• Achieved 90% file size reduction (to 30 KB) with inference times under 1 second.

• C++
• Java
• Microsoft SEAL
• OpenCV
• Dlib

Key Contributions

• Developed an object detection model to identify changes in website elements.
• Created a module to analyze the DOM and model outputs to determine which elements were modified.
• Reduced manual QA efforts by 50%.

• Python
• JavaScript
• Selenium
• Bootstrap
• YoloV5
• HTML5
• CSS3

Key Contributions

• Developed a module that utilizes a layout-based OCR model to extract information from post orders.

• Python
• PDF Plumber
• OpenCV
• Nanonets

Chess-AI

Developed with Django, the python-chess library, and chessboard.js, this web-based chess application allows users to participate in chess matches against an AI opponent. The bots are developed using different algorithms, with one utilizing minimax and another employing a Convolutional Neural Network trained on a kaggle chess dataset.

• Python
• JavaScript
• Pytorch
• chessboard.js
• python-chess
• ONNX Runtime

Flashcards

A Django-based web application designed for efficient review of solved leetcode questions. Users can input the problem link, specify a topic for categorization, and provide their solutions. The application will automatically generate dedicated pages for each problem by utilizing GraphQL queries to the leetcode/graphql server for getting additional question information.

• Python
• JavaScript
• GraphQL
• Bootstrap