# GradientWarrior

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Amit Binjola](https://cerebralvalley.ai/u/ammycodes)
- **GitHub:** https://github.com/ammycodes/aegis-edge
- **Demo video:** https://youtu.be/KYTNXoU1QDs
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2370

Scientific researchers spend hundreds of hours manually reviewing thousands of medical papers, often missing critical contradictions between studies—such as one paper finding that a drug inhibits a cellular pathway while another claims it activates it under different conditions.

Aegis-Edge solves this by acting as an automated evidence extraction and cross-study audit engine. Powered by the Meta Model API (muse-spark-1.3), Aegis-Edge converts unstructured scientific text (PubMed abstracts, PDFs, or manual text feeds) into validated knowledge triples (Subject ➔ Predicate ➔ Object), constructs an interactive knowledge network, automatically detects opposing biological claims, and provides cited, context-aware answers to complex research questions.

Key Features & Capabilities
📥 Multi-Modal & Batch Ingestion:

Flexible ingestion supporting direct text input, batch PubMed ID fetching via NCBI Entrez E-Utilities, and multi-file research PDF uploads parsed in real-time.

📋 Structured Knowledge Extraction:

Extracts precise biomedical relationships with confidence scores, exact verbatim evidence snippets, and finding statuses (SUPPORTED, CONTRADICTORY, CONDITIONAL, UNCERTAIN) backed by strict Pydantic schema validation.

⚠️ Automated Cross-Study Conflict Auditor:

Identifies opposing mechanisms across separate studies (e.g., INHIBITS vs. ACTIVATES for the same biological interaction) and flags them with real-time warning cards alongside detailed AI reasoning reports.

🕸️ Interactive Knowledge Graph Network:

Visualizes extracted entities and relationships using a dark-mode PyVis network graph, automatically highlighting disputed inter-study relationships in bright red.

💬 Evidence Synthesis & Scientific Q&A:

Allows researchers to query the loaded evidence base in natural language, generating context-aware summaries that explicitly cite supporting and conflicting sources (e.g., [study-pd1-alpha], [study-pd1-beta]).

## More from Meta Global AI Developer Hackathon

- [EnglishTalk](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2367)
- [Ram-Altman](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2368)
- [AxeCap](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2369)
- [LargitData — FoundFact](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2371)
- [Lautaro Pereyra](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2372)
- [Iris AI 2.0](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2373)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2370. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt