Zeeshan Akram Email & Phone Number

Pakistan

• Conduct R&D on Deep Reinforcement Learning. Develop architecture, simulation environment, algorithms, efficient and optimize solution code.• Design, implement, and optimize deep learning models for various use cases.• Write code for scalable training pipelines for machine learning models.• Write code for automated model training, validation, and deployment processes using MLOps best practices.• Create and maintain scripts for data preprocessing, model training, and evaluation.• Set up and manage cloud-based infrastructure (e.g., AWS) for model training and deployment.• Ensure code quality, reproducibility, and compliance with industry standards and best practices.• Document workflows, processes, and technical decisions clearly and comprehensively.

Islāmābād, Pakistan

• Fetch data from different sources. Perform data cleaning, preprocessing, wrangling and ETL.• Analyze data and get meaningful information and patterns. • Build machine learning and deep learning models and deploy them.• Perform experimentation and feature engineering to improve model accuracy.• Deploy entire AI software on the client site.• Developed Mlops from architecture level to implementation.• Use different AWS cloud resources for different tasks.• Build Python and bash scripts for automation.• Help the Mechanical team with different custom algorithm development.• Test, analyze and discuss custom algorithms performance with the Mechanical team and deploy on the production environment.• Agile Development.• Visit client's sites for on-premise deployments.

United States

• Participated in machine learning and deep learning competitions.1) M5 Forecasting - AccuracyThe objective of this challenge was to predict item sales at stores in various locations for two 28-day time periods.2) SIIM-ISIC Melanoma ClassificationThe challenge was to predict the probability (floating point) between 0.0 and 1.0 that the lesion in the image is malignant (the target). The model must be predicting a binary target for each image.3) ALASKA2 Image SteganalysisThe goal of the competition was to determine which of the images in the test set have hidden messages embedded.