Pediatric healthcare has a promising future with AI, and we got to experience this firsthand while engineering for our client, Heartful Sprout. It’s a clinical nutrition software platform that helps parents manage their child's nutrition, track new food introductions and get support for conditions ranging from celiac disease or diabetes to inflammatory bowel disease. As more families joined the platform, the work on the clinical side grew more and more complex.
While the app benefited parents greatly, behind the scenes, the dietitians did a lot of strenuous homework in order to serve parents. Each recipe required the team to spend hours on safe ingredient substitutions, nutritional calculations, preparation step adjustments and documenting the rationale behind every single change made to meet a patient's specific needs. As a result, recipe modification became one of the most time-consuming aspects of their clinical work. Ultimately, this challenge led the Heartful Sprout team to partner with GoML.
Why Heartful Sprout built the workflow before Gen AI
Before getting in touch with the GoML team, Heartful Sprout decided to pull AI completely out of the platform. The team wanted to watch how clinicians handled their days in actual practice. They had already put the main technical pieces in place like clinical dashboards, mobile apps and data pipelines. Yet they still did not have a solid sense of exactly where clinicians lost the most time.
Heartful Sprout stepped back and let the dietitians work through their normal manual routines. Along the way, the team tracked every step and noted the places where the work slowed down, turned repetitive or felt inconsistent. This preparation handed GoML a clear target once the work began. The focus stayed narrow on recipe modification rather than any broad AI addition.
Why GoML was the perfect partner for Heartful Sprout
Heartful Sprout ran its entire system on Amazon Web Services. AWS was also backing up the AI pilot through its generative AI program. This setup created clear expectations for any partner with proven skill with AI, solid experience on AWS and a firm grasp of healthcare regulations.
GoML met all three requirements and had already built AI solutions on AWS for other healthcare clients.
What was built into the clinical nutrition software
GoML built the recipe modification engine using the AI Matic Agentic AI blueprint and hosted it on AWS. The system extends Heartful Sprout’s existing clinical nutrition software without replacing any core parts. It runs on Amazon Bedrock, Claude Sonnet 3.7, FastAPI, Streamlit, PostgreSQL, and Amazon S3 with a vector knowledge base.
Heartful Sprout’s PostgreSQL database already held more than three thousand single-ingredient records, each tied to FDC ID numbers for nutritional analysis. GoML designed the engine to work inside this database rather than create a new one.
- Clinicians can now describe changes in plain language and type something like “Adjust this oat bran muffin recipe for a child with celiac disease”. The system reads the request, flags ingredients that must be removed due to restrictions, and suggests safe replacements that meet the needed criteria.
- The engine also pulls in clinical dietary guidelines for each condition. It applies rules around ingredient restrictions, suitable substitutes and preparation adjustments based on medical protocols, not general cooking knowledge. This allows it to separate options that are simply gluten-free from those that qualify as celiac-safe, respecting the clinical distinctions that matter.
- Every substitution stays linked to the original database records. Recipe titles, ingredient lists, and instructions keep their structure and formatting. Ingredient IDs remain fully intact, so nutrition values can still be recalculated later. In a clinical environment, this traceability was never optional.
Timeline and delivery of clinical nutrition software
GoML ran active development for approximately two weeks. At the end of that period, presented Heartful Sprout with a working demo. Heartful Sprout reviewed the output against how it would function in real clinical workflows and provided feedback, after which GoML completed one to two additional iterations before delivering a production-ready PoC.
Results
- 70-80% reduction in manual recipe adjustment effort
- 2x faster generation of disease-specific compliant recipes
- 50% improvement in dietary guideline consistency across the clinical team
- 100% ingredient ID integrity maintained for nutritional recalculation
Watch the complete discussion between Kay Lim (CEO and Founder, Heartful Sprout) and Rishabh Sood (Founder and CEO, GoML) in our full GoLive episode to learn more.
The picture of success with clinical nutrition software
Heartful Sprout is now helping over 500 families with tube feeding and complex condition management through its pediatric nutrition app for tracking and managing their children's care. Their long-term product roadmap includes supporting all types of pediatric healthcare providers in closing the technology gaps that exist today.
Stay tuned to the GoML blog to learn more about our 130+ production-grade developments in the world of AI and ML.
