Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

I’m Mahmoud Gamal Salem, an AI/ML engineer and researcher with a track record of building scalable language models and practical AI systems across finance, healthcare, and multilingual applications. I’m passionate about translating research into real-world impact, collaborating across teams, and continually improving model calibration, efficiency, and deployment for robust, responsible AI.…

Mahmoud Gamal Salem

AI Engineer, Data Scientist, Full Stack Developer, +2





I’m Mahmoud Gamal Salem, an AI/ML engineer and researcher with a track record of building scalable language models and practical AI systems across finance, healthcare, and multilingual applications. I’m passionate about translating research into real-world impact, collaborating across teams, and continually improving model calibration, efficiency, and deployment for robust, responsible AI.…

Available to hire

I’m Mahmoud Gamal Salem, an AI/ML engineer and researcher with a track record of building scalable language models and practical AI systems across finance, healthcare, and multilingual applications.

I’m passionate about translating research into real-world impact, collaborating across teams, and continually improving model calibration, efficiency, and deployment for robust, responsible AI.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Language

English

Fluent

Work Experience

AI/ML Engineer at Google

December 1, 2024 - Present

Developed ML systems to enhance real-world financial crime and money laundering detection workflows using LLMs and Agentic AI, driving measurable impact on fraud prevention. Led on-device audio generation to synthesize diverse cross-lingual data, enabling efficient adaptation and multilingual capabilities for on-device deployment.

Research Scientist at Cerebras Systems

December 1, 2024 - October 15, 2025

Developed and maintained end-to-end ML pipelines for large language models and foundational models supporting multilingual and multimodal applications. Led research on domain-specific biomedical LLMs (MediSwift) using 75% weight sparsity during pre-training on Cerebras CS-2 hardware, achieving 22.5x training efficiency and state-of-the-art results on PubMedQA and biomedical VQA. Worked on training Cerebras-LLaVA, a foundation vision-language model, on Cerebras hardware to advance multimodal understanding.

AI Research Intern at Google MTV

March 1, 2023 - October 15, 2025

Optimized language models using non-autoregressive generation and studied the effects of large-scale pretraining for non-autoregressive language generation. Published UT5: Pre-training non-autoregressive T5 with unrolled denoising.

Graduate Student Researcher at Vector Institute

March 1, 2023 - October 15, 2025

Supervisor: Prof. Graham Taylor. Worked on improving model calibration in computer vision problems under long-tailed distribution settings.

AI Research Intern at Borealis AI

February 1, 2022 - October 15, 2025

Modeling uncertainty using selective nets for classification and regression problems. Published Gumbel-Softmax Selective Networks at NeurIPS 2022 workshop.

AI Research Intern at Samsung AI Center Toronto

September 1, 2021 - October 15, 2025

Applied knowledge distillation in an unsupervised framework to leverage unlabeled and unpaired data across modalities. Worked on compressing multimodal transformers for vision and NLP using knowledge distillation.

AI Research Engineer at Navinfo Europe

August 1, 2020 - October 15, 2025

Explored adversarial robustness in semantic segmentation and object detection within multi-modal learning to improve generalization. Implemented network compression and distillation to meet real-time requirements (70% compression) and worked on real-time multi-object tracking on hardware (C++ & TensorFlow).

AI Research Intern at Valeo

August 1, 2018 - October 15, 2025

Optimized data annotation processes by 80% with pre-annotations; performed scene understanding using Camera/LiDAR fusion; researched and implemented state-of-the-art perception modules for autonomous driving (object detection, segmentation, tracking) and behavioral cloning in driving simulations with multi-task learning.