Avyantra: Leveraging ML to Rapidly and Accurately Diagnose Neonatal Sepsis (India)

Exploring challenges and opportunities for a startup in healthcare using ML for a more rapid and accurate diagnosis of neonatal sepsis

Avyantra Health Technologies is based in India with the mission to blend technology and innovation for enabling accessible and affordable healthcare in developing countries. Amongst their products, the focus of this blogpost will be PreSco – a machine learning platform for the early assessment of Neonatal Sepsis.

AI as a Service

Neonatal sepsis is a blood infection that occurs in infants younger than 90 days. Lack of a specific diagnostic test to detect neonatal sepsis though, prevents positive birth outcomes and timely treatment. The condition cannot be determined by one or two simple parameters. Avyantra has developed the product PreSco (Predictive Scoring) – which uses machine learning algorithms to provide an early risk assessment of Neonatal Sepsis and aids in prescribing antibiotics for the treatment plan. Through analyzing information on several neonate data points inputted by the healthcare professional, the platform generates a predictive score that helps doctors diagnose the condition. The score calculates the likelihood of an infant developing sepsis in the next 24 to 48 hours. This improves diagnostic accuracy and timeliness – facilitating early treatment for infants affected by this condition.

Status: What makes this company interesting to analyze is that it was founded in 2017 – and still in an early stage of its existence. It must grow rapidly as a startup and validate its product to scale and capture value, beyond just creating it. Key funders of the company include: UNICEF, the Indian incubator Villgro, and leading scientific institutions in India.

Advantages and Value Creation

  • Advantage over Current Methods of Detection: Current methods for detecting neonatal sepsis include blood testing (PCR) – wherein a small amount of blood is drawn. However, the volume of blood obtained, low presence of bacteria, and maternal antimicrobial infection are factors that make it very hard to accurately diagnose the condition. However, Avyantra’s solution provides a non-invasive, no-contact method for diagnosis which can provide rapid, real-time sensitive and specific diagnostics. According to leading medical journals, “Diagnostics with a faster turnaround time would not only improve surveillance in all settings but also facilitate timely management.”
  • Cloud-Based: The product is managed fully on the cloud – which means given newer offerings from AWS and Microsoft, the operating expenses of the technology itself can be kept low. Further, the cost of securing the data and maintaining data privacy can be reduced and the responsibility distributed. Hosting on platforms like AWS and Microsoft would enable them to double up the layers of security for their data and platform. It would also enable seamless transmission, availability, and analysis of data in real-time as well as scalability and efficient collaboration across stakeholders using this data for decision-making.
  • Open-Source: The technology behind the product is open source – in line with the mission of the company to make healthcare accessible and affordable to developing countries. According to the founders, the open source nature of the product also gives them the opportunity to share and gather key insights and experiences given the feedback from a vast, global community of data scientists and innovators. These insights will only help strengthen the solution as well as provide a low-cost way of testing, building, and deploying the product and future updates rapidly.

Challenges and Opportunities in Deployment

  • Acquiring Data: The product is still in stages of testing and validation – which means that a significant challenge for the company is to establish partnerships for acquiring large datasets from clinics, hospitals, and doctors. While health data regulations in India still lag far behind the US, given the vast and complicated healthcare network – this is going to be a significant logistical challenge.
  • Training Datasets and Potential for Bias: Developing countries are vastly diverse and different. The product is first being launched and tested in India – which is already a vastly diverse country. Ensuring that the datasets can be trained to support different cultural contexts within and across countries while minimizing bias will be an important engineering challenge.
  • Go-To-Market Strategy: Much like VideaHealth, Avyantra has several stakeholders in the healthcare space that it could sell this product to. Given that it is in its mature stages of validating its technology, this is going to be a critical decision that influences how it commercializes this technology. The company could follow a model like VideaHealth where it distributes this technology to neonatal and paediatrics departments in hospitals, or in paediatrics clinics, or to a different customer altogether like IGOs including the UN and governments.
  • Open Source Nature: While the open source nature of this product is expected to benefit the growth and refinement of the product itself, it could pose a significant challenge to business and commercial growth down the line (especially re. patenting). In such a situation, it would have to develop a clear profit formula as part of its business model and think about a ‘freemium’ model potentially – where a base source code providing the core services is open source and customizable add-ons can be offered to customers.


Despite a long list of challenges, there are also massive opportunities for the company to capitalize on – due to the dual bottom line nature of the product – which enables cost and time savings in healthcare and improved neonatal mortality outcomes for countries.

  • Unlocking New Types of Customers: There is a significant opportunity to unlock new types of customers like governments and international organizations – given the affordability of the product since it comes from a developing country.
  • Horizontal Expansion: Assuming that the company can access large datasets with several parameters indicating neonatal health outcomes, it can expand horizontally to provide value for other neonatal conditions as well. Building a large dataset can help train the model to provide early detection and diagnosis, personalized treatment plans, and other probabilities for health outcomes in infants that can help doctors choose the best course of treatment. Further, it will support traditional diagnostics instead of taking away from it – offering doctors a tool that helps improve the accuracy of other tools that they are currently using.
  • Organizational and Personnel: The founding team brings together strong skills and considerable experience in marketing, healthcare, IT, and analytics. The diversity of the team can be a key factor in ensuring that this product (and company) grows sustainably while improving processes and solutions – which will ultimately help bring a variety of perspectives for solving the problem of diagnosing neonatal sepsis.



Unilever: Using AI to freeze out competition & discover “breakfast for dessert”


Heuritech: AI for fashion trends

Student comments on Avyantra: Leveraging ML to Rapidly and Accurately Diagnose Neonatal Sepsis (India)

  1. Fantastic read Snigdha! Taking a step back, I was curious about how extremely specialized this approach seemed to be (eg, unlike what we saw with the dental case, this is focused on a single type of issue). While you mentioned horizontal expansion in their future, did you find any way they’re setting up the organization now to set themselves up for success?

  2. Thanks Snigda! I was interested to learn more about the “several neonate data points inputted by the healthcare professional” — do you know what these are in particular?

    I think there’s a interesting feedback look for AI systems, where at first they have to start with data that’s easy to gather. However, after the system is established, the promise of better predictiveness may itself be an incentive that motivates people to collect more costly data as an input to the system.

Leave a comment