BumpCare AI | Faster Reviews, Smarter Care, Safer Pregnancy

BumpCare AI speeds up prenatal care by summarizing records, flagging risks early, suggesting safe meds, and providing diet, exercise, and research insights — improving outcomes for mothers and babies.

Comment

Inspiration

Pregnancy is a critical phase where both the mother’s and baby’s health must be closely monitored. Doctors often juggle multiple reports — lab results, scans, vitals, and medical history — making it time-consuming to reach quick and safe decisions.

In resource-constrained hospitals, doctors and nurses may have very limited time for in-depth patient review. This increases the chances of oversight and delays in detecting complications.

Many digital systems today lack clinical intelligence and only present raw data without actionable insights or decision support.

We built BumpCare AI to act as a doctor’s daily assistant — helping clinicians save time, detect risks early, and deliver evidence-backed maternal care consistently.

What it does

BumpCare AI is a smart pregnancy-care assistant that:

  • Summarizes patient data trimester-wise for faster clinical review.
  • Detects complications like pre-eclampsia, anemia, thyroid disorders, and gestational diabetes.
  • Highlights similar cases from the database for comparative insights and improved decision-making.
  • Suggests non-branded medication classes, plus tailored food and exercise guidance.
  • Provides a doctor dashboard with alerts, trimester summaries, explainability features, and direct links to trusted research.
  • Ensures transparent outputs, where every AI recommendation is backed by citations from guidelines or peer-reviewed studies.

How we built it

  1. Data Handling – Structured patient data into trimester-based timelines (vitals, labs, scans, history) and linked them with guidelines (WHO, ACOG).

  2. AI Modules – Integrated PubMed-based retrieval and used FAISS for similarity search across patient data and research literature.

  3. Decision Support System – Built modules for summarization, medication class suggestions, nutrition/exercise guidance, and risk detection. Core reasoning, safe recommendations, and clinical evidence summaries were powered by Ollama gpt-oss-20b, ensuring explainability and citations.

  4. Interface – Designed a clean doctor dashboard with summaries, alerts, case comparisons, and explainability features.

  5. Clinical Evidence Integration – Embedded peer-reviewed research (Maternal Diet and Nutrient Requirements in Pregnancy, Daily Iron and Folic Acid Supplementation, Maternal and Child Nutrition) into the ontology for RAG-based recommendations.

  6. Safety Guards – Added value-range validation, unsafe drug filters, and enforced non-branded outputs to ensure clinical safety.

Tech Stack Overview

  • Frontend – HTML/CSS/JS → Login, Patient ID input, textual AI outputs
  • Backend – Flask → DBAuth, patient record aggregation, orchestrates AI
  • AI Models – Ollama gpt-oss-20b → Summaries, risk reasoning, safe recs, citations
  • Patient DB – SQL → Stores patient visits, labs, history
  • Vector Store – FAISS → Research, ontology, similar cases
  • Security – DB-Based Auth → Role-based access

Challenges we ran into

  • Data Diversity – Hospitals record data in inconsistent formats, making data normalization a major challenge.
  • Clinical Validation – AI recommendations had to strictly align with medical guidelines.
  • Trust Balance – Finding the right mix between speed of insights and transparent reasoning.
  • Governance & Privacy – Ensuring HIPAA/GDPR compliance without overcomplicating usability.
  • Edge Cases – Handling rare pregnancy conditions where limited reference data was available.

Accomplishments that we’re proud of

  • Built automated trimester-wise summaries that consolidate fragmented records into one clear view.
  • Implemented AI-driven risk detection & alerts for pre-eclampsia, gestational diabetes, thyroid disorders, fetal growth restriction, and anemia.
  • Developed an AI recommendation engine for safe medication classes, trimester-based nutrition, and lifestyle guidance.
  • Added case similarity analysis, enabling clinicians to learn from past comparable cases.
  • Integrated evidence-based medicine by linking AI suggestions directly to trusted clinical research.
  • Designed workflow support features that save doctor time and reduce cognitive overload.
  • Delivered nutrition & lifestyle recommendations: iron-rich foods, calcium/folate support, hydration, and safe trimester exercises (light yoga, walking, stretching).
  • Built in explainability features so clinicians can trace how AI reached a conclusion.

What we learned

  • Responsible AI in healthcare must be transparent, explainable, and clinically validated.
  • RAG pipelines (PubMed + FAISS) greatly improve evidence-based recommendations.
  • Structuring patient data by trimester accelerates clinical interpretation.
  • Clinicians prefer clear, actionable insights over opaque predictions.
  • UI/UX simplicity is critical — doctors favor clean, fast dashboards.
  • Multidisciplinary collaboration between AI engineers and clinicians is essential for adoption.

What’s next for BumpCare AI

  • Pilot with hospitals and integrate directly with EHR/EMR systems.
  • Expand beyond pregnancy into postpartum care and child health.
  • Scale similar-patient retrieval with larger, diverse datasets.
  • Add multimodal support — interpret lab reports, scans, images, and voice input.
  • Enable wearable integration (BP cuffs, glucose monitors, smartwatches) for real-time monitoring.
  • Extend multilingual support to rural and non-English-speaking regions.
  • Run clinical validation studies to measure time saved, accuracy, and patient outcomes.
  • Incorporate doctor feedback loops for continuous AI improvement.

Built With

+ 3 more
Share this project:

Updates