Stars
A practical template for structuring Claude Code projects with specialized subagents, persistent memory, and lifecycle hooks.
A scale-invariant atlas of head-specific, non-aligned manifolds in transformer attention residuals (O_h ≈ 0.28). Began as Fisher-eigenbasis quantization (NQP — refuted); refuting it revealed the ge…
CAL-L4: Meta-inference M(V) over compressed tensor volumes. L4 Efficiency Hypothesis · RCC · AMD-Instinct MI300X collaboration for O(n²) hardware baseline Topics: meta-inference, efficiency-hypothe…
Research framework for compressing multi-agent AI pipeline state into decision-relevant cognitive abstractions (L0–L4). Pre-paper: DOI 10.5281/zenodo.20430343 Topics: cognitive-oversight, multi-age…
CAL-L3: Composition operator C mapping cognitive tensor sessions T⁽ˢ⁾ to unified volume V. Tucker decomposition candidate · Governance Manifold Hypothesis · NeurIPS/ICML Topics: tensor-decompositio…
ROCm/HIP research filling gaps vs CUDA on AMD Instinct MI300X — wave-agnostic primitives, DME-pipelined Flash Attention, topology-aware Infinity Fabric AllReduce, and MLIR MFMA auto-tiling.
ResearchLab es un entorno integrado para equipos de investigación científica y académica. Gestione proyectos, hipótesis, notas, bitácoras y colaboradores en un solo lugar, potenciado por inteligenc…
Virtual Security Officer (VSO): Responde preguntas de usuarios (clientes o interesados). Usa documentación interna como base de conocimiento Genera respuestas con un modelo de lenguaje (LLM). Restr…
Carbohydrate calculator by aliments recognition with deep learning throug neural network and artificial intelligence
Endo-Edge LAB+ Sistema ciberfísico de diagnóstico temprano y seguimiento continuo de enfermedades crónicas
TCO-L2: Human orchestration of AI pipelines via cognitive tensor T[d,i,j,k] and the Natural Cognitive Frontier (NCF). RCT n=40 · CHI 2027 · CAL Layer 2 Topics: cognitive-oversight, human-ai-interac…
Repositorio de Prácticas, ejercicios y proyectos dictados en el curso de Harvard CS50W