Lezwon Castelino
Machine Learning Engineer at Swym
Old Toronto, Ontario, Canada
Summary
🤩
Rockstar🎓
Top SchoolLezwon Castelino is a Machine Learning Engineer with 11 years of software engineering experience, currently working at Swym in Toronto. He brings strong systems-level skills in C++, Ruby, and Python combined with applied ML expertise from meaningful contributions to high-profile open-source projects like pytorch/vision and PyTorch Lightning. His work on datasets, model ports, TPU integration, and ONNX export reflects a focus on making research-ready models production-friendly and hardware-aware. Lezwon has a Master of Computer Applications from Christ University and a track record of teaching and mentoring, having developed a beginner Python course for BigBinary Academy. Comfortable across the full ML stack, he blends low-level performance tweaks with end-to-end training and deployment considerations. His background suggests a pragmatic engineer who translates complex model and infrastructure challenges into reliable, user-focused solutions.
11 years of coding experience
4 years of employment as a software developer
Christ University, Bangalore
Stackoverflow
55reputation
2kreached
1answer
4questions
Github Skills (24)
pytorch10
python10
machine-learning10
pytorch-lightning10
tpu10
datasets10
deep-learning10
computer-vision10
model-driven9
testing9
onnx9
modeling9
model-driven-development9
model-building9
flash-memory6
Programming languages (9)
Github contributions (5)
Lightning-AI/pytorch-lightning
Apr 2020 - Mar 2021
Pretrain, finetune ANY AI model of ANY size on multiple GPUs, TPUs with zero code changes.
Role in this project:
ML Engineer
ML EngineerContributions:72 reviews, 19 commits, 27 PRs in 10 months
Contributions summary:Lezwon primarily contributed to enhancements and bug fixes related to the integration and support of TPU (Tensor Processing Unit) hardware within the PyTorch Lightning framework. Their work included modifications to the trainer, distribution parts, and documentation to enable users to select individual TPU cores, configure mixed-precision training, and address issues related to saving and loading models on TPUs. The contributions also involved refactoring tests and adding features, such as ONNX export compatibility to facilitate model deployment.
pythonheadachespytorch-modelsdata-sciencehandling
pytorch/vision
Mar 2022 - Jun 2022
Datasets, Transforms and Models specific to Computer Vision
Role in this project:ML Engineer
ML EngineerContributions:19 reviews, 5 commits, 6 PRs in 2 months
Contributions summary:Lezwon primarily contributed to the `pytorch/vision` repository by implementing and testing new datasets, specifically the USPS dataset, and modifying existing datasets such as Kinetics. They also worked on integrating multi-weight support and porting various models like Alexnet, ConvNext, and DenseNet. The user's contributions include bug fixes and improvements related to the models and the data loading pipelines.
pytorchvisiondeep-learningdatasetcomputer-vision
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.