I had the opportunity to present recent work Leveraging Domain Knowledge For Deep Learning Based Computer Vision at Knowledge Graph Conference 2022. The blurb for this talk is as following:

Sometime early this year, I deployed a brand new thoroughly tested, and profiled application onto a self-managed Kubernetes cluster. Suffice to say, all hell broke loose. The pods were getting OOMKilled with error code 137 left and right. This sparked a massive crime scene investigation and some interesting insights were discovered. I blogged about this in May, 2021. I have made some more progress on this since then, as a sidekick.

I spoke about our recent experience working with ML pipelines at Kafka Summit 2021. The topic was Building More Reliable Data Pipelines for Nearmap's Deep Learning Models: An Evolutionary Case Study link.

Had the priviledge on being invited to speak at AWS Community Day 2020 to talk about deep learning data curation:

Incredibly excited to be speaking at AWS 2020’s She Builds Day event!

Nearmap captures terabytes of aerial imagery daily. With the introduction of AI capabilities, Nearmap has leveraged Kubernetes to generate AI content based on tens of petabytes of images effectively and efficiently.

Nearmap captures terabytes of aerial imagery daily. With the introduction of artificial intelligence (AI) capabilities, Nearmap has leveraged Kubernetes to generate AI content based on tens of petabytes of images effectively and efficiently.

Creating reproducible machine learning system is hard! It is because challenges pertaining to reproducibility stem from all 5 ingredients of machine learning systems - software, hardware, algorithms, process and practice and data.

Reproducibility crisis is real in data-science. This crisis has been recognized in data-science research and several efforts e.g. International Conference on Learning Representations has been underway in improving reproducibility in data science research. Dr. Joelle Pineau, an Associate Professor at McGill University and lead for Facebook’s Artificial Intelligence Research lab (FAIR) in her talk covered the importance of reproducibility.

Significant amount of effort has been invested in improving the quality of breast imaging modalities (for example, mammography) to increase the accuracy of breast cancer detection. Despite that, about 4-34% of cancers are still missed during mammographic examination of cancer of the breast. This indicates the need to explore a) The features of the lesions that are missed, and b) Whether the features of missed cancers contribute to why some cancers are not ‘looked at’ (search error) whereas others are ‘looked at’ but still not reported. In this visual search study, we perform feature analysis of all lesions that were missed by at least one participating radiologist.

At Nearmap, we have been working on an open source CI/CD solution for Kubernetes based workload called KCD for Kubernetes Continous Delivery (Formally known as CVManager (Container Version Manager)). We had oppertunity to present this solution at KubeCon EU Copenhagen, Denmark 2018. For more details of this talk see link and associated blog with detailed info can be found here.

Visual search, the process of detecting and identifying objects using the eye movements (saccades) and the foveal vision, has been studied for identification of root causes of errors in the interpretation of mammography. The aim of this study is to model visual search behaviour of radiologists and their interpretation of mammograms using deep machine learning approaches. Our model is based on a deep convolutional neural network, a biologically-inspired multilayer perceptron that simulates the visual cortex, and is reinforced with transfer learning techniques.

The rapid evolution in medical imaging has led to an increased number of recurrent trials, primarily to ensure that the efficacy of new imaging techniques is known. The cost associated with time and resources in conducting such trials is usually high.