AI-Powered Photo Tool Has Doubled Food Tracking Engagement: Healthify CTO

For health and fitness app operator Healthify, the incorporation of GenAI has led to a better connection with its user base. Speaking at Inc42’s ‘The GenAI Summit’ today, the startup’s CTO Abhijit Khasnis said that Healthify’s AI-powered food recognition system has enabled the company to double user engagement in food tracking.

“With a snap, we can see that people who switch to tracking using photos do twice the tracks than what people do manually. So, it’s a 2X higher engagement, which means there is a 2X higher impact that we are seeing with those users,” he explained during a panel discussion focussed on AI’s impact on the healthcare sector.

The health and fitness startup launched the AI-powered feature in 2023 to recognise Indian food from images for logging in calorie intake, thereby allowing users to track their meal intake more efficiently. Under this, when users upload food photos to the app, the system uses vision language models to identify food items and proprietary algorithms to understand user preferences.

The GenAI-based recognition system was launched to enhance its chatbot ‘Ria’ and build ‘Snap’, which helps users count the calories in their meals. In 2024, the startup had claimed that GenAI usage resulted in a 50% increase in food tracking, deeper user engagement, and 18% higher client-coach interaction.

“At the core of it is the vision language models, where it’s looking at the picture and then saying, ‘I can see these foods in it’. Then we use heuristics and our own proprietary understanding of what a user is likely to eat to track it for the user,” Khasnis said.

The increased tracking directly correlates with better health outcomes, as Khasnis noted that users who track more consistently see better results in their health goals. “We see a direct correlation between people who track more, lose more weight.”

The Bengaluru-based healthtech startup, which raised $20 Mn last October to close its pre-Series D round at $45 Mn, has been using AI to improve user experience and health outcomes. The startup then claimed that it achieved operational profitability in India and was on track to become EBITDA-positive by the end of FY25.

Khasnis was joined by Chaitanya Raju, chief product officer at HealthPlix, and Laina Emmanuel, CEO and cofounder of BrainSight AI, to discuss and answer the question, “AI In Healthtech: Redefining Care Or Just Optimising It?”

Trust A Major Challenge In AI Adoption In Healthcare 

Building trust remains the most critical challenge for AI applications in healthcare, with each panelist sharing different approaches to the problem.

“It took us five years to get our software as a medical device licence and a whole bunch of tests to be able to show the accuracy,” Emmanuel explained. “We used to sit down with doctors after every case and explain the science, the papers behind what we were doing.”

BrainSightAI, which recently secured $5 Mn, creates functional brain maps that help surgeons identify critical areas to avoid during operations. 

The startup validates its technology during awake brain surgeries, where surgeons can directly confirm if the AI’s predictions match what they observe when stimulating different brain regions. 

Healthify’s Khasnis noted that trust develops through hands-on experience with the technology. “For us, trust comes through experience,” he said, describing how users often test the system themselves by taking photos of their meals and then removing items to see if the AI correctly identifies the changes.

Talking about other challenges and new developments, Khasnis warned about what he calls “AI blindness,” where users dismiss AI-generated content without consideration. 

Meanwhile, Raju said that HealthPlix is planning to expand its voice-to-text feature moving forward to simplify medical record creation for doctors.

“One of the things we have done with GenAI in the last six months is we have launched what we call a voice-to-text capability, where all the doctor has to do is turn on the mic, and the model automatically converts the conversation into a summary,” said Raju.

The panelists agreed that AI’s role in healthcare should remain assistive rather than diagnostic. “It’s important for us to understand what AI can do and what it cannot do. No model in the world can diagnose well enough right now,” Raju said.