Technology & Research

AI in service of healing

BCF is pioneering the use of artificial intelligence to carry out nonprofit health missions — making world-class pediatric research accessible to every family, in every language, for free.

Why AI, Why Now

Three forces converge.

The moment for AI-powered pediatric health intelligence is not coming — it is here. And it is passing the families who need it most.

Force One

The knowledge gap is measurable and devastating.

A family in suburban Texas just received a Crohn's diagnosis for their 12-year-old. They turn to Reddit and Facebook groups because the next specialist appointment is three months away. And the wait is just the beginning — families navigate years of treatment decisions, medication side effects, and evolving diagnoses with no trusted, research-backed guide. Over 100,000 American children under 20 live with IBD — with Crohn's disease prevalence rising 22% in the last decade. The research exists. The guidance does not reach the family.

BCF exists to close this gap.
Force Two

The pediatric chronic disease burden is global and growing.

In India, childhood cancer claims tens of thousands of young lives annually in regions where a pediatric oncologist is hundreds of miles away. Thalassemia affects an estimated 100,000 children born every year in India — the highest burden globally. These families have no specialist to call and no trusted source of guidance in their language. Technology is not a replacement for physicians — but it is a force multiplier for the knowledge those physicians would provide.

100,000+ new pediatric Cancer and Thalassemia patients annually in India alone.
Force Three

The AI moment has arrived — but not for everyone.

Large language models have fundamentally changed what is possible in health information delivery. But the families who need this most — navigating a new diagnosis alone at 2am, searching Reddit for answers that belong in a clinic — are precisely those being left behind by English-only, account-required, data-harvesting commercial platforms.

BCF is building the alternative: privacy-preserving, multilingual, free forever.
How We Use AI

The research loop that powers BCF.

Three interconnected systems — not three products. One mission expressed in three dimensions.

COMPASS Research Corpus PathLight Family Interface CANOPY Intelligence Layer BCF MISSION

Research signals feed COMPASS. COMPASS powers PathLight. PathLight serves families. Families generate signals that inform BCF's research grant decisions. Funding enables better research.

This is not three products — it is one mission expressed in three dimensions. BCF's design intent is a fully closed loop: family queries reveal gaps, gaps direct enrichment, enrichment improves answers, usage signals inform grant decisions, grants fund better research, better research enters the corpus. The first two links are operational today. The remaining links close as BCF scales — each one strengthening every other.

Our Applications

Built for families who have no one else to ask.

BCF's platform closes the knowledge gap between world-class research and the families who need it most.

Live Now

PathLight

Research-backed guidance for families navigating pediatric chronic disease. Anonymous. No account required. Powered by 288,000+ and growing research papers and AWS Bedrock. Available in 9 languages including Spanish.

Hindi Marathi Tamil Telugu Bengali Gujarati Kannada English
Visit pathlight.beejcare.org →
Coming 2027

Rural Health Worker Assistant

Millions of community health workers serve families in underserved regions worldwide — each reaching hundreds of families with limited access to specialists. BCF is building a voice-first, multilingual assistant for these workers — powered by Amazon Polly and AWS Bedrock. One trained health worker equals one thousand families served.

Voice-first Community Health Multilingual AWS Bedrock
Learn about COMPASS →

COMPASS — Research Intelligence Engine

COMPASS indexes, enriches, and semantically searches 288,000+ and growing pediatric chronic disease papers from the world's leading scientific databases. Every paper is processed through BCF's AURA enrichment pipeline — extracting 12 AI-enriched fields using local LLM inference.

COMPASS powers PathLight's real-time semantic search — matching family questions to the most relevant research findings in milliseconds, across 7 conditions and 9 languages.

288,000+
Papers indexed
384
Dim. embeddings
32
Fields per paper
5
Daily data sources
AURA Enrichment Pipeline
1
Daily Ingest
New papers pulled from PubMed, OpenAlex, bioRxiv, medRxiv, Semantic Scholar.
2
AI Enrichment
Each paper processed by Qwen 32B — extracting key findings, plain-language summaries, and 9 additional fields.
3
Vector Embedding
384-dimensional semantic embeddings generated, enabling similarity search across the full corpus.
4
pgvector Storage
HNSW index enables sub-millisecond semantic search at scale on AWS EC2.
5
PathLight Delivery
Top 5 papers retrieved → synthesized by Bedrock Claude → delivered in family's language.
Data Sources

Where the research comes from

PubMed / MEDLINE
National Library of Medicine — gold standard for biomedical literature.
OpenAlex
Open-access index of global scholarly research — 250M+ works.
bioRxiv / medRxiv
Preprint servers for cutting-edge research — frontier signals.
PubMed Central (PMC)
Full-text open access archive of life sciences literature.
Semantic Scholar
AI-powered research tool from Allen Institute — 200M+ papers.
AURA Enrichment

What we extract from every paper

Key FindingMain result in one sentence
Plain LanguageParent-readable summary
Study TypeRCT, review, meta-analysis...
Patient PopulationPediatric / adult / mixed
Age RangeAges studied
TreatmentMain intervention
Outcome SignalPositive / negative / mixed
Research StageDiscovery → clinical
GeographyCountry or region
Novelty FlagCutting-edge indicator
Family Relevance0.0–1.0 score
Frontier SignalResearch frontier tag
How We Build

A new model for nonprofit technology

BCF is pioneering the use of AI to carry out nonprofit health missions in ways that have never been done before.

BCF operates on a Human-in-the-Loop (HITL) development model — where human judgment drives every architectural decision, and AI executes with precision. This model gives BCF the technical velocity of a well-funded startup at nonprofit cost discipline.

SEVA — Strategic AI Advisor

SEVA (Sanskrit for "selfless service") is BCF's AI strategic advisor — powered by Claude (Anthropic). SEVA architects systems, designs products, drafts strategy documents, and helps BCF operate with the velocity of a well-funded startup while maintaining the cost discipline of a nonprofit.

KARMA — AI Development Partner

KARMA is BCF's AI coding partner — powered by Claude Code (Anthropic). KARMA builds, tests, and deploys BCF's technology under human oversight. The pattern: SEVA plans → Board decides → KARMA executes. All critical decisions remain with BCF's human board.

Next Evolution — Agentic Development Model

BCF's current HITL model has one constraint — conversational AI assistance loses session context across long sessions and carries disproportionately higher token costs combined with Claude Code. BCF's next development model: Claude Code paired with Cline and a local Model Context Protocol (MCP) server, using an open-source LLM as a local context engine. This upgrade multiplies development velocity while reducing per-feature token cost — every saved dollar flows directly to the families BCF exists to serve.