Open
to anybody with an concept
Microsoft for Startups Founders Hub brings folks, data and advantages collectively to assist founders at each stage resolve startup challenges. Enroll in minutes with no funding required.
Have you ever ever observed a spot in your pores and skin and puzzled if it could possibly be one thing severe? Puzzled in case your cough indicated COVID or simply the run of the mill flu? Want you might verify your blood strain along with your smartphone? Enter Helfie.ai, an rising healthcare startup within the healthcare sector who’re harnessing multli-modal AI algorithms designed to assist sufferers shortly and precisely verify for a spread of healthcare circumstances. With groups in Australia and Denmark, Helfie.ai goals to make healthcare extra accessible and inexpensive by enabling sufferers to investigate their well being circumstances from the consolation of their properties, utilizing solely their smartphone.
Helfie.ai is constructing an answer that caters to healthcare suppliers who can arm their sufferers with an utility to extra shortly and precisely entry the care they want Helfie.ai’s AI mannequin analyzes footage, movies, sound recordings, together with face scans by way of a sufferers’ cellular gadget to verify for indicators of potential well being circumstances. A number of the diagnoses they help are respiratory well being, vitals, and pores and skin most cancers. The product helps real-time physician help to verify the prognosis and prescribe the suitable remedy.
Helfie.ai is a part of the Microsoft for Startups Pegasus program, which gives entry to Azure credit, Go-to-Market (GTM), technical help, and distinctive advantages similar to entry to Azure AI infrastructure on a devoted GPU cluster.
Helfie.ai is leveraging a devoted NDm A100 v4-series cluster in Azure ML to coach and finetune the AI mannequin that energy their resolution. Due to the necessity for medical grade predictions a couple of affected person’s well being situation, their flagship product, Very important’s AI, analyzes movies of sufferers faces to guage well being circumstances and make distant assessments.
By utilizing the devoted GPU cluster of A100s, Helfie.ai’s CTO Nikhil Sehgal, studies that by utilizing the devoted GPU cluster of A100s, they’ve diminished their mannequin coaching time from two weeks to simply two hours.
“The drastic discount in mannequin coaching time is essential for a startup like us, it permits us to construct quicker and get our product into the palms of customers,” Sehgal tells Microsoft for Startups.
The GPU cluster additionally permits them to scale up their fashions as they develop their consumer base and information sources, making certain that their merchandise can deal with the rising demand and complexity. Operating workloads on Azure additionally helps them seamlessly adjust to compliance insurance policies just like the Well being Insurance coverage Portability and Accountability Act (HIPAA) and Normal Knowledge Safety Regulation (GDPR).
Helfie.ai has developed specialised AI fashions, often called “AI Specialists,” fine-tuned for particular medical domains. As an illustration, the “Cardiology AI” is an LLM (Massive Language Mannequin) with complete data in cardiology, enabling customers to obtain expert-level responses much like these from an expert heart specialist. Helfie.ai accomplishes this by accumulating respected literature and documentation from the net to fine-tune each open-source LLMs and flagship GPT-series fashions.
To create and monitor these fashions, Helfie.ai makes use of Azure ML Studio for machine studying experiments, hyperparameter optimization, and efficiency reporting. The corporate depends on Azure Kubernetes Service (AKS) for intensive computational duties, Azure OpenAI for multimodal LLM capabilities, and Azure AutoML for optimizing hyperparameters. They ran each classification and laptop imaginative and prescient mannequin experiments on Azure ML. Via these experiments, Helfie.ai was capable of optimize hyperparameters together with superb n_estimators and max_depth.
Helfie’s assessments typically contain visible or audio inputs, demanding exceptional computational bandwidth to coach their workloads. A major problem was re-training the pores and skin most cancers prediction mannequin utilizing smartphone photos of lesions. To deal with this, Helfie.ai established its complete coaching pipeline on Azure’s devoted GPU cluster, enhancing throughput per epoch as new information from educational establishments necessitated frequent re-training.
To deal with information privateness, Helfie implements sturdy AI content material security insurance policies to make sure consumer interactions with specialist LLMs are secure and stay confidential. This consists of Azure AI Content material Security Dashboard, a real-time dashboard that flags responses with potential dangers, serving to to forestall any dangerous consumer actions.
Helfie.ai’s merchandise—Vitals AI, Respiratory AI, STI AI, and Skin AI—are actually accessible on Azure Market.
Devoted GPU clusters can be found for startups on Microsoft for Startups who’re backed by considered one of Microsoft’s strategic VC companions: Microsoft expands free Azure AI infrastructure access to startups | Microsoft for Startups Blog, together with Startups who’re lively within the Pegasus Program.
