APURBA TECHNOLOGIES LIMITED Software Engineer - Machine Learning
Dhaka, Bangladesh | May 2023 - Present

• Design, implement, and maintain computer vision and NLP models for various applications, such as image and text classification, object detection, and named entity recognition.
• Pre-process, clean, train and validate computer vision and NLP models using large datasets, optimizing parameters and algorithms to improve accuracy and performance.
• Deploy computer vision and NLP models into production systems and integrate them with other systems and applications.
• Manage and maintain the infrastructure required for ML model training and deployment, including cloud services, computing resources, and data storage.
• Stay up to date with the latest research in computer vision and NLP and perform experiments to test new algorithms and approaches.
• Identify new opportunities to apply computer vision (2D & 3D), NLP (Multilingual), and Machine Learning to different projects.

WEBNATION GLOBAL LIMITED Full Stack Software Engineer
Nagoya, Japan (remote) | January 2023 - May 2023

• Developed web-based applications for different client projects using PHP/Laravel, JavaScript, MySQL, jQuery, VSCode, and AJAX.
• Participated in the design, testing, and code implementation according to software design specifications, by following standard coding styles and practices.
• Wrote clean and maintainable code and documentation that is easy to understand and modify.
• Conducted code reviews to ensure code quality and consistency.
• Mentored and guided other team members.
• Collaborated closely with cross-functional teams to identify business requirements and design scalable and maintainable solutions.

APPLIED QUANTITATIVE SCIENCES, INC. Junior Data Scientist
Fort Lauderdale, FL, USA (remote) | May 2022 - December 2022

• Created detailed report with PowerBI on optimizing work items for Abbott’s equipment servicing team which predicted to decrease the response time by at least 15%.
• Defined Product Strategy and Development Roadmap for Pavement Condition Profiler Web Application.
• Developing a Minimum Viable Product on Asphalt Condition Monitoring for Florida Department of Transportation aiming to drastically reduce inspection cost and increase inspection intervals.
• Automated routine Key Performance Indicator analysis and assessment for Johnson & Johnson’s procurement team.
• Developed an application to report the load status of server nodes using InfluxDB and Python for Data Visualization allowing us to implement a load balancer to reduce cloud server costs by 10%.

METAGEEK, LLC. Software Engineer
Boise, ID, USA (remote) | May 2021 - April 2022

• Created variations of existing web pages with Nuxt.js for running split testing experiments using Google Tag Manager and Google Optimize.
• Analyzed site engagement data from Google Analytics and made data-driven decisions to create scalable web pages.
• Rebuilt and Optimized SEO for Educational and Product pages in modern Vue.js-based architecture (Nuxt.js) resulting in a 23% increase in overall traffic.
• Worked closely with designers to take wireframes and mockups from conception to implementation on Nuxt.js web framework.
• Participated in daily stand-up meetings, mid-sprint check-in meetings, and sprint planning meetings to plan and execute sprint tasks.

SMART INFRASTRUCTURE RESEARCH LAB, BSU Research Assistant
Boise, ID, USA | May 2017 - May 2019

• Trained a neural network model with TensorFlow based on Single Shot Multibox Detector architecture for pavement surface crack detection.
• Achieved 96% prediction accuracy for detecting pavement cracks.
• Designed and built a small-scale autonomous vehicle platform equipped with Nvidia Jetson TX2 for data collection and real-time inference for pavement crack detection.
• Developed an android application that communicates with SCiO near infra-red spectroscope via Bluetooth and sends collected data to a cloud server for analysis.
• Experimented with various methods of classifying different grades of asphalt samples and created a machine learning classification model that identified grades and amount of additive in asphalt with an average accuracy of 85%.