Why South Asia Is the Most Important Battlefield for Medical AI

Dengue kills between 10,000 and 20,000 people per year. The WHO estimates 390 million dengue infections occur annually, 96% of them in tropical and subtropical regions. Sri Lanka, India, Bangladesh, Pakistan — these are the epicenters.

There are no FDA-approved AI diagnostic tools specifically designed for dengue severity prediction. There are no widely deployed machine learning models trained on South Asian clinical populations for thalassemia variant classification. There is no publicly available AI tool built for the specific presentation patterns of malaria in the South Asian monsoon belt.

This is not because the problem is too hard. It is because the problem is too far from where the money is.

Where medical AI investment goes

In 2023, NIH spent $7.4 billion on AI-related medical research. The majority of that investment went to cardiovascular disease, Alzheimer's, and oncology — conditions with high prevalence in the United States and Western Europe.

Dengue research received $47 million. Thalassemia AI research received $12 million. Malaria AI, despite being the second-largest infectious disease killer globally, received $89 million — less than 1.2% of what was spent on Alzheimer's AI alone.

Google DeepMind's AlphaFold is a genuine scientific breakthrough. It has transformed protein structure prediction. The drugs it will help discover will primarily target diseases that kill Western populations, because that is where the pharmaceutical market is.

This is not a criticism of DeepMind. It is an observation of incentive structures.

The South Asian disease burden

The diseases disproportionately affecting South Asia share a set of characteristics that make them tractable for AI:

• Large patient populations generating substantial clinical data
• Relatively constrained disease mechanisms compared to complex neurodegenerative conditions
• Diagnostic bottlenecks solvable by pattern recognition — exactly what machine learning does well
• Severe shortage of trained diagnosticians in rural areas — exactly where AI-assisted diagnosis has the most leverage

Dengue severity prediction is fundamentally a classification problem. Clinical data including platelet count trajectories, NS1 antigen levels, and fever curve patterns have established predictive relationships with severe dengue outcomes. A model trained on South Asian hospital data — not transferred from a Western dataset — could meaningfully assist triage decisions in underfunded clinics.

What ASI Research Lab is building

Monolith 02 starts with three disease areas: dengue, thalassemia, and malaria. The selection criteria were:

1. Disease is disproportionately prevalent in South Asia
2. Diagnostic data is available or obtainable via hospital partnerships
3. AI-assisted diagnosis or severity prediction is technically feasible within 24 months
4. No equivalent tool exists for South Asian clinical contexts

Data partnership discussions are active with hospital networks in Sri Lanka. The first published research output target is December 2026. The goal is not an academic exercise — it is a deployable diagnostic support tool that a clinic in rural Sri Lanka can use.

The world's largest AI labs are not building this. That is why we are.