DTM E27. How to Build AI Products – Ankit, Qure.ai
From the Deep Tech Musings Podcast - Get actionable and tactical insights to take your Deep Tech startup from 0 to 1 [Idea to Traction]
Listen now on - Spotify, Apple, Google
Ankit is a Senior Product Manager and Founding member of Qure.ai, a healthcare AI startup. He lends insights into how they started on their path to finding the right problem to solve, how they assessed the AI problem fit, and the technical feasibility of it in the early days. Ankit also shares his input about the technology behind the product and how they build their MVP. Finally, he emphasizes the important point on how to go about building AI teams and also gives some advice for aspiring AI product managers.
Listen to the episode here,
or on the below platforms
Where to find us:
Ankit Modi – Anki Modi (LinkedIn)
Qure.ai - Qure.ai (LinkedIn)
Pronojit Saha, DTM Podcast - pronojitsaha (LinkedIn), @pronojits (Twitter)
Show Notes & Summary:
(2:49) Founding story of Qure.ai - from data struggles to initial success
Ankit had the idea of using AI in healthcare but was open to finding the right area to make an impact. After market research and talking to clinicians, they realized that radiology was a suitable field for AI intervention.
Initially, they faced a challenge in acquiring data to create algorithms as hospitals and radiology companies required proof of their AI credentials.
To overcome this, they selected a set of problems (example: brain tumors, head CT scans, etc.) with publicly available datasets and worked on them for several months.
Once they had something to showcase, they approached the companies again and gained access to data, allowing them to start building their first algorithm.
(5:14) On business metrics & the role of the “board of customers”
Ankit identified key metrics that could be improved through AI in healthcare, such as reducing the average reporting time for radiology exams and decreasing the turnaround time for reports.
The importance of early partners or customers, referred to as a "board of customers," was highlighted in defining problems and metrics, as they played a role in shaping the company's direction and providing valuable feedback.
In B2B scenarios, feedback from a smaller number of customers is crucial, and Ankit expressed gratitude towards their initial partners for their patience and assistance in improving the product.
(13:40) How to assess technical feasibility of an AI solution?
Technical feasibility is crucial after developing an AI algorithm, focusing on making it consumable and seamlessly integrated into existing workflows.
Ankit mentioned the importance of interpretability, highlighting the need to explain why AI makes certain recommendations or decisions.
They built applications for medical teams, such as multidisciplinary teams for lung cancer, to collaborate and make informed decisions based on radiological exams.
Ankit described a case in the capital city of the Philippines where their qXR solution was deployed. By integrating the solution into a van that collected X-rays, they reduced the diagnosis cycle for TB from four weeks to just a couple of hours.
The impact of AI in this case included identifying 25% more cases of TB, preventing the spread of the disease, and achieving a highly satisfying and impactful outcome.
(18:40) From research to deployment: Navigating the uncertain path of AI product development
AI is viewed as a tool for creating impactful products, but the uncertainty associated with research and development (R&D) requires product managers to account for that.
A typical AI product team includes members from R&D, design, engineering, and, in the case of B2B products like healthcare, operations and sales, as well as a regulatory team for obtaining clearances.
AI research is considered more scientific, while deploying AI models is seen as more engineering-oriented, necessitating different teams to handle these aspects.
During the early stages of Qure, Ankit and another founding member wrote a Python app to productionize their first algorithm and deployed it in a hospital. However, as scalability became a factor, engineering challenges arose, emphasizing the need for skilled engineers to take the product from initial stages to a larger scale.
(25:05) Ankit’s overall approach to building AI products
The first version of the product was a simple API where radiology images could be uploaded and the API would detect a few findings on the image.
The early version allowed doctors to confirm whether the detected findings were accurate by returning a replica of the image with marked findings.
Building AI products in healthcare is still in an experimental stage, and being nimble and able to experiment quickly is crucial.
Clear communication within the team and consensus on the target quality of the product are important aspects of the development approach.
Scaling was facilitated by learning from deploying the qXR product and expanding into other modalities based on customer feedback and suggestions.
(34:13) Ankit’s advice for aspiring AI product managers
AI should be seen as a tool to solve customer pain points, and product managers should focus on the customer's needs.
Transitioning from tech to product management requires identifying blind spots and working on them, such as improving design skills.
Having a technical foundation, including knowledge of machine learning, is beneficial for an AI product manager to collaborate effectively with the development team.
A product manager should have a good understanding of various disciplines involved in building a product, including back-end development, front-end development, design, and sales.
Regular sessions between technical and non-technical team members can promote cross-learning and improve understanding of AI and its deployability in non-technical terms.
If you enjoyed this episode, please leave us a rating on Spotify, Apple, or wherever you listen to podcasts. It helps us reach more people who are interested in deep tech & grow the community.
Also, don’t forget to subscribe to Deep Tech Musings Podcast on pronojits.substack.com so you never miss an episode.
Thank you for listening! See you next time!