🏆 For Judges / Prize tracks alignment
🏆 Prize Track Alignment

One project. Five tracks.

The Gemma 4 Good Hackathon offers a Main Track ($50K 1st prize) and four Special Technology Prizes ($10K each), plus a LiteRT Special Tech Prize ($10K). Per Section 2.5 of the rules, projects are eligible for both a Main Prize AND a Special Tech Prize. JemmaPass is engineered for maximum overlap.

💰 $50K Main · $10K Special 🎯 Max payout: $60K 📋 Rules §2.5
🥇

Main Track · 1st Prize

The grand prize for the most impactful Gemma 4 Good project

$50,000
Judging criteria: Impact & Vision (40 pts) · Video Storytelling (30 pts) · Technical Depth (30 pts). The project that best fuses all three wins.

Why JemmaPass wins this track. Three factors compound. First, the impact surface is genuinely universal — pilgrims, travelers, refugees, disaster volunteers, rural elderly. Second, the video narrative spans Belgium 🇧🇪 and Japan 🇯🇵 with a Hollywood-grade dramatic arc: peaceful pilgrimage → earthquake → SOS → killer Augmentin moment → life saved. Third, the technical depth is genuinely novel: nobody else fuses Gemma 4 + HL7 IPS R4 + Nearby Mesh into a single offline-native Android artifact.

IMPACT 3 healthcare gaps closed simultaneously: connectivity, semantic, sovereignty. Real-world origin in Noto 2024 + Shikoku × 3.
VIDEO 3-min Hollywood arc. 18 scenes. Killer 02:21 red alert moment with anaphylaxis-averted denouement.
TECH DEPTH 49 native @Tool functions · 3.36 GB on-device KB · 16 KB ELF aligned · 302K DDI records.
🏥

Health & Sciences Special Prize

Best project advancing clinical, biomedical, or scientific applications

$10,000
Why this track exists: Healthcare AI is a major use case for Gemma 4. This prize honors the project that delivers the deepest, safest, most clinically rigorous medical AI integration on-device.

Why JemmaPass wins this track. This is the clearest fit. Three production-grade clinical features distinguish JemmaPass from chatbot-grade health AI:

  • Zero-translation paradigm
  • ATC-hierarchy-aware allergy cross-reactivity
  • DDInter 2.0 fully embedded
DDInter 2.0 302,516 DDIs · 2,310 drugs · 8,398 mechanisms (NAR 2024)
HL7 IPS R4 v2.0.0 STU 2 standard · 18 pillars · zero translation needed
SNOMED CT IPS Free Set + 25-language refset · July 2024 release
UMLS 2025AB ~1.4M filtered concept rows · NLM Nov 2025
ATC/DDD 2026 ~6,900 codes · 5-level hierarchy · WHO Collaborating Centre
RxNorm April 2026 Drug strengths, dosage forms · NLM April 6, 2026
💡 Regulatory positioning. JemmaPass is positioned as a decision-support tool for emergency resilience and patient empowerment, NOT a medical device under EU MDR Class IIa SaMD or FDA SaMD classification. This is intentional: it allows the project to ship without years of regulatory clearance, while still providing genuine clinical value in resource-constrained settings.
🌍

Global Resilience Special Prize

Best project building resilience to disasters, climate, or system failures

$10,000
Why this track exists: Edge-based AI is uniquely positioned to deliver critical services when traditional infrastructure fails. This prize honors the project that turns the connectivity weakness of disasters into an opportunity for on-device intelligence.

Why JemmaPass wins this track. Disaster resilience is the founding premise.

NOTO 2024 Author was disaster volunteer · M 7.6 JMA · 280+ deaths in immediate aftermath
P2P_CLUSTER Mesh topology M-to-N · BLE + Wi-Fi Direct · zero cellular dependency
MULTI-HOP RELAY 6 chunk types · VR/SR carry TTL · profiles propagate through chain of phones
OFFLINE TTS Android TTS engine local · alerts spoken even when network is gone
PAPER FALLBACK Pocket Pass PDF printable · works when the device itself is destroyed
NFC HCE Host Card Emulation · victim phone advertises as a medical card · no pairing
🌐

Digital Equity Special Prize

Best project advancing access to AI for underserved populations

$10,000
Why this track exists: AI assistance should not be a luxury reserved for English-speaking tech-fluent populations with broadband. This prize honors the project that makes on-device AI genuinely accessible across language, infrastructure, and class divides.

Digital equity at every layer.

UI LANGS 15 BCP-47 languages · catalog in IpsLanguageCatalog.kt
KB LANGS 25+ languages × IPS terminology codes in ips_valuesets_translations
NO SUBSCRIPTION MIT-licensed during hackathon → CC-BY 4.0 on prize award
NO ACCOUNT Zero PII collected · zero telemetry · zero analytics
PAPER FIRST Pocket Pass PDF printable · printer-friendly A4 layout
VOICE INPUT Whisper-class STT + Gemma 4 audio · works for illiterate users
⚙️

LiteRT Special Technology Prize

Best project showcasing the LiteRT on-device runtime

$10,000
Why this track exists: LiteRT (formerly TensorFlow Lite) is Google's on-device inference runtime. The LiteRT-LM extension specifically supports Gemma 4's structured tool calling. This prize honors the project that most deeply exercises LiteRT's capabilities.

Deepest LiteRT integration in the field.

LiteRT-LM v0.11.0 (or latest available) · litertlm-android:0.10.1
Gemma 4 E2B 2.4 GB model · supported for older / lower-RAM Pixel devices
Gemma 4 E4B 3.4 GB model · default on Pixel 9 Pro XL
@Tool count 68 native function-calling tools (24+18+17+4+3+2)
Multimodal Text + image + audio (via RIFF/WAVE wrapper)
16 KB aligned All .so files verified align 2**14 on Pixel 9
📊 Scoring summary

Maximum cumulative payout: $60,000

Per Section 2.5 of the Competition Rules, a single project can win 1 Main Prize + 1 Special Technology Prize. Below is the realistic payout matrix.

Combo Likelihood Payout
🥇 Main 1st + 🏥 Health & Sciences Best fit — clinical depth is the strongest moat $60,000
🥇 Main 1st + ⚙️ LiteRT Strong technical fit — 49 @Tool is unmatched $60,000
🥇 Main 1st + 🌍 Global Resilience Story fit — Noto 2024 origin is compelling $60,000
🥈 Main 2nd + 🏥 Health & Sciences Solid fallback if main is highly contested $35,000
🏥 Health & Sciences (only) Conservative floor — clinical depth alone $10,000

Your verification tour continues

You've seen the architecture, the data layer, and the prize alignment. Reproducibility, benchmarks, FAQ, and team coming in the next deliveries.

← For Judges hub ⚙️ Architecture 💊 DDInter