The workshop
Capability listing for evaluators, integrators, and engineers. Each line is what the platform does today, validated on the canonical USSOCOM TE 26-2 corpus or on the engineering bench.
01 / 02
Capabilities
Hypha ingests, profiles, and reasons about operator data without a notebook. The exploratory pass surfaces distributions, anomalies, target leakage candidates, and group-stratification risk before the model is built.
VALIDATED / TE 26-2 CORPUSModel Building Agent runs feature engineering, model family selection, training, and evaluation end to end. The canonical run completes in 2 hours 13 minutes wall clock on a 48 GB workstation.
VALIDATED / 2:13 WALL CLOCKGroup-stratified holdouts. Memorization detection. DO NOT DEPLOY verdicts surfaced in the operator UI with auto-generated recovery guidance. The credibility lever.
VALIDATED / HOLDOUT DISCIPLINESemantic Fusion Engine merges real-time data streams with operator-supplied doctrine and reference material. Predictive synthesis with citations the commander can audit.
VALIDATED / CITED SYNTHESISOn-disk ChromaDB. In-memory BM25. Approximately 16,000 retrieval chunks on the canonical corpus. Citations baked into every retrieval. No commercial cloud LLM.
VALIDATED / ~16,000 CHUNKSVisualization Agent composes alerts, heatmaps, time-scrub, and simulations from the model output. The dashboard the commander reads.
VALIDATED / ENGINEERING BENCH02 / 02
The test floor
Every capability above is held by a test suite. The bench runs on every change. The number is on this page because it is the number on the bench, not because it is a marketing claim.
Backend tests
8,595passing
Frontend tests
784passing
Canonical end-to-end
2:13wall clock
Reference hardware
48GB workstation