The full capability breakdown, per-GPU benchmarks, and the private Blackwell cluster spec — for the technical buyer who wants to see under the hood.
Closed-loop automation over EVV systems, scheduling, payroll exceptions, and onboarding. Local-LLM-only, encrypted state, full audit trail.
Continued pretraining over your private documents, then domain SFT, then retrieval grounding. We train on our cluster — your data never sees a third-party API.
A purpose-built agent framework — declarative agent contracts, unified tool registry, per-agent budget + full audit log, PHI-aware model routing, and a single operator review inbox. 235B-parameter MoE primary with fallback paths.
Owns the envelope-to-signature loop: DocuSign orchestration, personalized chase, escalation triggers, human-escalation paths.
Voice cloning over minutes of reference audio, real-time lip-sync rendering, sub-second targets for interactive use cases.
Pipeline-built corpus ingestion (PDFs, regulatory text, treatises), vector retrieval, and a citation-verifier that hard-flags fabricated cites before they reach your user.
Benchmarked end-to-end on the actual cluster, not extrapolated from vendor decks. PyTorch microbench at 8K square matmul, BF16 + FP8 tensor-core paths, median of 200 iterations.
| Node | GPU | BF16 TFLOPS | % peak | FP8 TFLOPS | % peak |
|---|---|---|---|---|---|
| Spark 1 | GB10 | 92 | 74% | 184 | 73% |
| Spark 2 | GB10 | 91 | 73% | 182 | 73% |
| Node 3 | PRO 6000 Full | 356 | 71% | 668 | 67% |
| Node 3 | PRO 6000 Max-Q | 246 | 56% | 474 | 54% |
| Node 4 | PRO 6000 Max-Q | 245 | 56% | 524 | 60% |
| Node 4 | PRO 6000 Max-Q | 261 | 59% | 486 | 55% |
| Cluster total | 1,291 | 63% | 2,518 | 61% | |