AI researcher exploring reasoning, knowledge systems, agents, and scientific discovery.
I build systems that aim to move beyond pattern matching toward structured understanding, falsifiable hypotheses, and verification-driven research.
Can AI move from generating plausible answers toward discovering missing relationships, producing falsifiable hypotheses, and learning from verification?
Perception
↓
Structured representation
↓
Knowledge graph / geometry
↓
Assumptions and contradictions
↓
Unexplored relationships
↓
Hypotheses
↓
Falsifiable tests
↓
Validation and replication
| Area | Repository | Purpose |
|---|---|---|
| Discovery | Geometric-Engine-Intelligence- | Knowledge graph and unexplored-edge discovery prototype |
| Discovery | HONET- | Contradiction and assumption-driven research framework |
| Reasoning | Generalised-Meta-Attention-Architecture | Experimental reasoning and attention architecture |
| Adaptive reasoning | SUPHAI_MODEL | Adaptive inference and uncertainty experiments |
| Geometry | GeoSemAlign-Visual-Geometry-Symbol-Meaning-Pipeline | Geometry, symbols, and semantic alignment |
| Agents | symbiote-agent-live | Agent execution experiments |
| Collective reasoning | Synthetic-Reason-Collective-Intelligence-System | Multi-system reasoning experiments |
| Knowledge | Aurora-The-Conversational-Knowledge-Lens | Conversational knowledge exploration |
Contributions are most valuable when they test an idea rather than merely endorse it.
High-value work includes:
- independent replications;
- stronger baselines;
- ablation studies;
- mathematical formalization;
- counterexamples and falsification tests;
- benchmark datasets;
- provenance and evidence improvements;
- reproducibility fixes;
- alternative implementations.
A negative result is valuable when it is reproducible.
Claims should be separated into:
- Algorithmic novelty — the system produces a different or useful result.
- Knowledge novelty — the result is not already present in the evaluated corpus.
- Scientific discovery — the result survives empirical testing and independent replication.
The goal is to make the third claim difficult to earn and therefore meaningful when it is achieved.
I am interested in rigorous, testable work at the intersection of:
- AI reasoning and planning;
- knowledge representation and graph reasoning;
- geometry and symbolic structure;
- agentic systems and verification;
- reproducible scientific evaluation.
For shared contribution guidelines, see the project contribution guidelines.