Svastha means "Health" in Sanskrit. Svastha Intelligence translates to "Health Intelligence". By health, we mean biomedical/medicine and anything related to healthcare.
Svastha Intelligence is a research ideas page. The primary focus of this page is to:
In this proposal, we talk about how we utilized data science and machine learning to analyze, prevent and predict system errors in a dental milling device. The total market size estimated for the dental milling industry is close to $3billion+. Based on estimates by the World Health Organization, oral diseases affect close to 3.5 billion people worldwide. Often times these machines have a high cost of manufacturing and the unfavorable reimbursement policy hinders the customer experience. In this scenario, using data science/machine learning/artificial intelligence for predictive maintenance is absolutely critical. Based on years of data accumulated, we built systems that are capable of analyzing, preventing, and predicting system errors commonly found in a dental lab milling workflow. We show how we utilized state-of-the-art Natural Language Processing techniques and Classification algorithms to build a system that is able to analyze, prevent and predict system errors. We show how we efficiently applied data science techniques to aid in value generation and how we utilized state-of-the-art techniques in a manner that was applicable to a business problem. This talk is suitable for the general data science community and particularly data scientists/machine learning engineers interested to see how data science can efficiently be applied in a Computed Aided Manufacturing industry to yield favorable business outcomes.
At Swiggy’s Core-Logistics data science team which focuses on algorithms to improve the delivery experience, this is a doomsday scenario we are all afraid of! Swiggy has been growing at a remarkable pace since its inception. With the recent foray into grocery delivery, the growth does not seem likely to slow down anytime soon! Doomsday hasn’t happened yet, and we are working proactively on ways to avoid it! The solutions have been quite fun — like combining word-embeddings with maps-data to create location-clusters. This may sound like a random juxtaposition of ML buzzwords. But the solution we’ll walk you through in the upcoming blog does precisely this. First, let’s start with a brief overview of the assignment algorithm itself.
Biomedical text understanding and retrieval is an active field of research. Advances in genome sequencing technologies have led to an exponential increase in genome sequence data deposited in the public repository, such as the NCBI Sequence Read Archive (SRA). Despite having a vast amount of publicly available data for researchers to share and analyze, the existing NCBI SRA search capability is limited since it lacks the inclusion of semantics in its search functionality. Recent advances in Natural Language Processing (NLP) and Information Retrieval (IR) have allowed us to build search systems capable of understanding search terms' semantics. The influx of publicly available data and NLP advances such as Bidirectional Encoder Representations from Transformers (BERT) has allowed us to fine-tune our model on applicable downstream tasks such as text classification and Named Entity Recognition (NER). Here we present a search tool, SRASearch, that utilizes NLP techniques to improve search functionality from the NCBI SRA.