Independent internal evaluation against the ISED Canada TRL 1–9 framework. Scores are derived from source code review, test results (201 tests, 0 failures), integration evidence, and deployment state as of v0.4.0 GA. No score reflects roadmap capability.
DeployIQ's core infrastructure stack — event bus, storage layer, API server, agent framework, and SHA-256 audit chain — is integrated and verified against live components in a controlled lab environment (Docker, real databases, chaos tests). The differentiating AI layer (PPO inference) and GPU topology validation are designed and implemented but not yet validated in an operational environment with real GPU cluster hardware. The overall system is demo-ready and suitable for design partner evaluation. A single production deployment on NeoCloud hardware would advance the overall rating to TRL 6.
Each major subsystem assessed independently. The overall TRL is bounded by the lowest-TRL differentiating component — in this case the PPO/ML inference layer and GPU topology validation, both at TRL 4.
| File / Artefact | Assessment Point | Status | ||
|---|---|---|---|---|
| tests/integration/api_routes.rs | 25 API route integration tests against live Docker infra | ⊙ confirmed | ||
| tests/integration/storage.rs | 11 storage integration tests (TimescaleDB, Redis, Neo4j) | ⊙ confirmed | ||
| crates/deployiq-agents/src/tests.rs | 165 agent unit tests | ⊙ confirmed | ||
| crates/deployiq-agents/src/ml_adapter.rs | ACL to FastAPI ML server · MlPredictRequest / MlPredictResponse wire types | ⊙ confirmed | ||
| crates/deployiq-agents/src/agent_trait.rs | learn() is a no-op in default trait implementation | ⚠ no-op | ||
| crates/deployiq-agents/src/topology_validator.rs | NVLink / IB / RDMA validation logic — 330 lines, synthetic data | ◑ partial | ||
| crates/deployiq-agents/src/finops_agent.rs | Cost attribution model — calls local ML server, not provider billing APIs | ◑ modeled | ||
| config/deployiq-demo.toml | All endpoints localhost · demo API key · ml.server_url localhost:8090 | · design-partner mode | ||
| Cargo.toml workspace.package.version | Workspace version "0.1.0" vs marketed "v0.4.0 GA" — naming gap, not a bug | ◑ noted | ||
| deploy/docker-compose.test.yml | 4-container test stack · NATS, Redis, TimescaleDB, Neo4j · all ports mapped | ⊙ confirmed | ||
| Chaos test suite (5 scenarios) | NATS leader kill, agent offline, spot interruption, DRP tamper, GPU node failure — all passed | ⊙ confirmed | ||
| [feature][hybrid-compute] XPU support layer — extend DeployIQ agents to Intel Ga | roadmap | TRL 1 | Issue open — concept defined | #1354 |
| [feature][quantum] Quantum-safe API gateway — enforce PQC handshakes on all Depl | roadmap | TRL 1 | Issue open — concept defined | #1353 |
| [feature][ai-infra] AI infrastructure market intelligence dashboard — unified vi | roadmap | TRL 1 | Issue open — concept defined | #1352 |
| [feature][quantum] Quantum noise characterisation agent — tomography-based error | roadmap | TRL 1 | Issue open — concept defined | #1351 |
| [feature][ai-infra] GPU utilisation efficiency monitor — detect underutilised re | roadmap | TRL 1 | Issue open — concept defined | #1350 |
| [feature][quantum] Quantum-classical gradient computation bridge — VQE/QAOA vari | roadmap | TRL 1 | Issue open — concept defined | #1349 |
| [feature][ai-infra] Multi-provider SLA enforcement layer — validate uptime, late | roadmap | TRL 1 | Issue open — concept defined | #1348 |
| [feature][quantum] Quantum-classical QUBO solver bridge — dispatch combinatorial | roadmap | TRL 1 | Issue open — concept defined | #1347 |
| [feature][ai-infra] Cross-cloud egress cost optimizer — minimise data transfer c | roadmap | TRL 1 | Issue open — concept defined | #1346 |
| [feature][ai-infra] InfiniBand / RoCE fabric health monitor — integrate IB diagn | roadmap | TRL 1 | Issue open — concept defined | #1345 |
| [feature][ai-infra] Spot/reserved/on-demand cost-blending optimizer — minimise t | roadmap | TRL 1 | Issue open — concept defined | #1344 |
| [feature][hybrid-compute] Cross-compute FinOps — unified cost attribution across | roadmap | TRL 1 | Issue open — concept defined | #1343 |
| [feature][quantum] QPU cold-start prediction — estimate QPU initialisation laten | roadmap | TRL 1 | Issue open — concept defined | #1342 |
| [feature][quantum] Quantum virtual machine deployment verifier — validate QVM im | roadmap | TRL 1 | Issue open — concept defined | #1341 |
| [feature][quantum] Quantum OS and SDK compatibility checker — Qiskit / Cirq / Pe | roadmap | TRL 1 | Issue open — concept defined | #1340 |
| [feature][quantum] Quantum key distribution (QKD) integration — route cryptograp | roadmap | TRL 1 | Issue open — concept defined | #1339 |
| [feature][quantum] Post-quantum cryptography migration advisor — detect classica | roadmap | TRL 1 | Issue open — concept defined | #1338 |
| [feature][quantum] Quantum network topology validator — extend TopologyValidator | roadmap | TRL 1 | Issue open — concept defined | #1337 |
| [feature][quantum] Quantum volume and CLOPS benchmarking agent — continuous QPU | roadmap | TRL 1 | Issue open — concept defined | #1336 |
| [feature][quantum] Quantum circuit compilation cost estimator — gate count, dept | roadmap | TRL 1 | Issue open — concept defined | #1335 |
| [feature][quantum] Hybrid CPU/GPU/QPU workload decomposer — identify quantum-ame | roadmap | TRL 1 | Issue open — concept defined | #1334 |
| [feature][quantum] Quantum error correction overhead estimator — QEC cost model | roadmap | TRL 1 | Issue open — concept defined | #1333 |
| [feature][quantum] QPU availability agent — ORAL lifecycle agent for IBM/IonQ/Qu | roadmap | TRL 1 | Issue open — concept defined | #1332 |
| [feature][hybrid-compute] Inference cost-per-token normalizer — $/1M tokens acro | roadmap | TRL 4 | Issue open — concept defined | #1331 |
| [feature][network] BGP-aware workload placement — route training jobs to regions | roadmap | TRL 3 | Issue open — concept defined | #1330 |
| [feature][hybrid-compute] Cross-provider GPU workload scheduler — dynamic dispat | roadmap | TRL 3 | Issue open — concept defined | #1329 |
| [feature][ai-infra] Contract lock-in advisor — recommend capacity pre-commitment | roadmap | TRL 3 | Issue open — concept defined | #1328 |
| [feature][ai-infra] Provider dependency risk scorer — flag Nvidia lock-in and de | roadmap | TRL 3 | Issue open — concept defined | #1327 |
| [feature][ai-infra] Compute backlog forecasting — predict capacity tightening be | roadmap | TRL 1 | Issue open — concept defined | #1326 |
| [feature][ai-infra] GPU generation placement optimizer — A100/H100/B200/GB200 ro | roadmap | TRL 1 | Issue open — concept defined | #1325 |
| [feature][ai-infra] Real-time multi-provider GPU compute price signal monitor | roadmap | TRL 1 | Issue open — concept defined | #1324 |
| IPGeoAI: Transformer-Based Geolocation with LLM Semantic Fusion | roadmap | TRL 1 | Issue open — concept defined | #1404 |
| From Answers to Interpretations: Rethinking Ambiguity-Induced Aleatoric… | roadmap | TRL 1 | Issue open — concept defined | #1403 |
| MaxKernel: Agentic Kernel Generation for TPUs | roadmap | TRL 1 | Issue open — concept defined | #1402 |
| Corporate Language Model (CLM): Transforming Tacit and Fragmented Enterprise… | roadmap | TRL 1 | Issue open — concept defined | #1401 |
| QMClaw: A Scalable General-purpose Framework for Quantum Measurement and Control | roadmap | TRL 1 | Issue open — concept defined | #1400 |
| A Sim-to-Real Study of Surface-Code Decoder Benchmarking | roadmap | TRL 1 | Issue open — concept defined | #1399 |
| Efficient Quantum Error Correction from Three Dimensional Qubit Control | roadmap | TRL 1 | Issue open — concept defined | #1398 |
| New codes aim to link quantum modules despite unreliable connections | roadmap | TRL 1 | Issue open — concept defined | #1397 |
| How Do Photonic Quantum Computers Work? | roadmap | TRL 1 | Issue open — concept defined | #1396 |
| The Steane Code Explained | roadmap | TRL 1 | Issue open — concept defined | #1395 |
| AI Is Forcing Data Centers To Rethink Trust | roadmap | TRL 1 | Issue open — concept defined | #1394 |
| Building Technology People Can Trust | roadmap | TRL 1 | Issue open — concept defined | #1393 |
| IBM Built the Cold War’s Most Powerful Code Breaker for the NSA | roadmap | TRL 1 | Issue open — concept defined | #1392 |
| The E.U.’s AI Drive Undermines Its Own Chip Strategy | roadmap | TRL 1 | Issue open — concept defined | #1391 |
| Everpure sees AI infrastructure strategy move beyond virtual machines | roadmap | TRL 1 | Issue open — concept defined | #1390 |
| Lightedge puts workload fit at the center of hybrid cloud optionality | roadmap | TRL 1 | Issue open — concept defined | #1389 |
| VMware Intros Private AI Cloud, AI Factory As Workloads Shift To On-Prem | roadmap | TRL 1 | Issue open — concept defined | #1388 |
| Show HN: LLMKube – Kubernetes for Local LLMs with GPU Acceleration | roadmap | TRL 1 | Issue open — concept defined | #1418 |
| Can LLMs in Draft-Verify-Revise Pipelines Resolve Deictic Ambiguity? | roadmap | TRL 1 | Issue open — concept defined | #1417 |
| Security Implications of Measurement Based Crosstalk on Superconducting Quantum… | roadmap | TRL 1 | Issue open — concept defined | #1416 |
| Quantum neural network equipped with backpropagation on a qudit processor | roadmap | TRL 1 | Issue open — concept defined | #1415 |
| Sparse-Blossom Decoding in $o(1)$ Time | roadmap | TRL 1 | Issue open — concept defined | #1414 |
| Parallel Circuit Execution for Scalable Quantum Computation | roadmap | TRL 1 | Issue open — concept defined | #1413 |
| Benchmarking the computational power of quantum computers | roadmap | TRL 1 | Issue open — concept defined | #1412 |
| Quantum cryogenic tests move from research labs into chip fabrication | roadmap | TRL 1 | Issue open — concept defined | #1411 |
| Heterogenous QEC codes boost efficiency in Quantinuum’s Helix architecture | roadmap | TRL 1 | Issue open — concept defined | #1410 |
| Heterogeneous Memory Chiplets Accelerate Multi-Request LLM Inference (NUS) | roadmap | TRL 1 | Issue open — concept defined | #1409 |
| Chip Industry Technical Paper Roundup: Sept. 14 | roadmap | TRL 1 | Issue open — concept defined | #1408 |
| Sponsored: Gerchamp launches modular AI data center to bring high-density… | roadmap | TRL 1 | Issue open — concept defined | #1407 |
| d-Matrix Joins the NVIDIA NVLink Fusion Platform | roadmap | TRL 1 | Issue open — concept defined | #1406 |
| Broadcom Rides Rocketing Trend For Custom AI Accelerators | roadmap | TRL 1 | Issue open — concept defined | #1405 |
| Shock-Capturing Quantum Algorithm for the Linear Advection Operator | roadmap | TRL 1 | Issue open — concept defined | #1458 |
| ReFINE: Scheduling of Distillation and Coding for Rate-Fidelity Tradeoff in… | roadmap | TRL 1 | Issue open — concept defined | #1457 |
| Hybrid quantum-classical attention for histopathology-based molecular profiling… | roadmap | TRL 1 | Issue open — concept defined | #1456 |
| Set-Packing and Sequence-Pair QUBOs for the 2D Cutting Stock Problem on Quantum… | roadmap | TRL 1 | Issue open — concept defined | #1455 |
| Researchers Build Colour Codes with Polynomial Error Correction | roadmap | TRL 1 | Issue open — concept defined | #1454 |
| Quantum Zeitgeist Weekly Digest | roadmap | TRL 1 | Issue open — concept defined | #1453 |
| BSC and iPronics connect AI applications to optical connectivity | roadmap | TRL 1 | Issue open — concept defined | #1452 |
| RSA-896 Has Been Factored. Steve Weis reports factoring RSA-896 using Claude | roadmap | TRL 1 | Issue open — concept defined | #1451 |
| Researchers Bound Condition Number for Faster Parity Learning | roadmap | TRL 1 | Issue open — concept defined | #1450 |
| AI in Chip Design: From Code Generation to EDA Orchestration (University of… | roadmap | TRL 1 | Issue open — concept defined | #1449 |
| Chiplet Co-Design Framework Reduces Energy and Design Costs for AI Accelerators… | roadmap | TRL 1 | Issue open — concept defined | #1448 |
| The AI Inference Revolution Is Here | roadmap | TRL 1 | Issue open — concept defined | #1447 |
| What to expect during CoreWeave’s ‘Fully Connected’ | roadmap | TRL 1 | Issue open — concept defined | #1446 |
| Astera Labs Releases Leo 2 CXL Memory Controllers and Leo X Controller for… | roadmap | TRL 1 | Issue open — concept defined | #1445 |
| Infleqtion And Nvidia Reduce Physical-to-Logical Qubit Ratio in Quantum… | roadmap | TRL 1 | Issue open — concept defined | #1444 |
| SMP: A General Hyperedge-Based Framework for Circuit-Level Quantum Error… | roadmap | TRL 1 | Issue open — concept defined | #1473 |
| High-Rate Quantum Codes with Proven Distance and Low-Weight Measurements | roadmap | TRL 1 | Issue open — concept defined | #1472 |
| Quantum computing’s “dark horse” just proved it can go universal | roadmap | TRL 1 | Issue open — concept defined | #1471 |
| Quantinuum and UWA Partner to Advance Australia’s Quantum Capability | roadmap | TRL 1 | Issue open — concept defined | #1470 |
| Chip Industry Week In Review | roadmap | TRL 1 | Issue open — concept defined | #1469 |
| HBF for High-Throughput LLM Serving (UC Berkeley, FuriosaAI) | roadmap | TRL 1 | Issue open — concept defined | #1468 |
| Extreme Environments Push Chips To Their Breaking Point | roadmap | TRL 1 | Issue open — concept defined | #1467 |
| Unveiling IC-STAR: Full-Flow Autonomy from Digital to Analog | roadmap | TRL 1 | Issue open — concept defined | #1466 |
| General Compute signs multi-year agreement with Cerebras | roadmap | TRL 1 | Issue open — concept defined | #1465 |
| Sponsored: Overcoming five common direct liquid cooling specification challenges | roadmap | TRL 1 | Issue open — concept defined | #1464 |
| AI cloud startup Verda raises $189m in Series B funding round | roadmap | TRL 1 | Issue open — concept defined | #1463 |
| Germany’s Karlsruhe Institute of Technology unveils 33 petaflops HoreKa 2… | roadmap | TRL 1 | Issue open — concept defined | #1462 |
| The modern F1 pit crew: AI engineers who race the deadline, not the car | roadmap | TRL 1 | Issue open — concept defined | #1461 |
| Touring the F5 BIG-IP Next for Kubernetes Lab to Make AI Clusters More Efficient | roadmap | TRL 1 | Issue open — concept defined | #1460 |
| Improving Quantum Error Correction By Meshing Surface Code With IBM’s Heavy-Hex… | roadmap | TRL 1 | Issue open — concept defined | #1459 |
| Mitigating Capability Degradation in LLM Self-Improvement via Representation… | roadmap | TRL 1 | Issue open — concept defined | #1488 |
| VEPA: Verified Evolving Formal Policy for Lifelong LLM Agent Guardrails | roadmap | TRL 1 | Issue open — concept defined | #1487 |
| EvoBelief: Falsifiable World Models for Self-Improving LLM Agents | roadmap | TRL 1 | Issue open — concept defined | #1486 |
| Faithful and Complete LLM-Assisted Automated Construction of Hierarchical… | roadmap | TRL 1 | Issue open — concept defined | #1485 |
| CLEAT: Learning Value-Guided Clarification for Interactive Agents | roadmap | TRL 1 | Issue open — concept defined | #1484 |
| Quantum-assisted rendezvous on hypercubes: a quantum Bernoulli noise approach | roadmap | TRL 1 | Issue open — concept defined | #1483 |
| Trainable Quantum-Native Language Models: Dissipative Memory, Capacity Limits,… | roadmap | TRL 1 | Issue open — concept defined | #1482 |
| Fast computational primitives for La-cross qLDPC codes | roadmap | TRL 1 | Issue open — concept defined | #1481 |
| Error correction and quantum information spreading on sparse random graphs | roadmap | TRL 1 | Issue open — concept defined | #1480 |
| TLS loss in superconducting resonators on epitaxial hBN thin films | roadmap | TRL 1 | Issue open — concept defined | #1479 |
| Supercomputing researchers document evolution of AI hardware | roadmap | TRL 1 | Issue open — concept defined | #1478 |
| Chip Industry Week In Review | roadmap | TRL 1 | Issue open — concept defined | #1477 |
| Master AI Chip Principles With New IEEE Design Program | roadmap | TRL 1 | Issue open — concept defined | #1476 |
| Nebius signs LOI with Volt for data center capacity in Rotterdam, the… | roadmap | TRL 1 | Issue open — concept defined | #1475 |
| IBM connects enterprise AI orchestration to production readiness ahead of… | roadmap | TRL 1 | Issue open — concept defined | #1474 |
| Concrete Functions in Private Simultaneous Quantum Messages: Quantum Advantages… | roadmap | TRL 1 | Issue open — concept defined | #1495 |
| Encodings and Node Architecture: Lessons from the Fault-Tolerant Distributed… | roadmap | TRL 1 | Issue open — concept defined | #1494 |
| Quantum-Assisted Unambiguous Discrimination of Pure States with Robustness… | roadmap | TRL 1 | Issue open — concept defined | #1493 |
| Chip Industry Week In Review | roadmap | TRL 1 | Issue open — concept defined | #1492 |
| Simulation App Speeds Up Electric Truck Battery Design | roadmap | TRL 1 | Issue open — concept defined | #1491 |
| Edge AI for Smart Cities: From Insight to Cognitive Action | roadmap | TRL 1 | Issue open — concept defined | #1490 |
| Chip legend Jim Keller’s Tenstorrent raises $1.5B to expand its open… | roadmap | TRL 1 | Issue open — concept defined | #1489 |
The event bus (NATS Raft), storage layer (TimescaleDB + Neo4j + Redis), API server, and SHA-256 DRP chain are integration-tested against live containers, chaos-tested to failure, and fully documented. This is the level of infrastructure rigor normally seen at Series B, not first design-partner evaluation. It significantly de-risks the deployment risk a NeoCloud partner takes on.
The current system calls a local FastAPI ML server that returns an action + confidence. The agent framework is architected for online RL (EpisodeTransition, RewardWeights, Learner trait) but the learning loop is a no-op in the base implementation. To a VC or technical buyer who reads Rust source: this is a rule-augmented inference system, not a continuously training PPO agent. This is fine for TRL 4–5 — but the benchmark's language ("PPO RL consensus") should be qualified in direct technical conversations to avoid credibility risk. The architecture for PPO is real; the trained production model is the gap.
topology_validator.rs correctly models the 5-layer GPU fabric stack (NVLink, IB, RDMA, SHARP, NUMA) with a well-structured 8-dimensional state vector. The validation logic is sound and chaos-testable. What it has not done yet is ingest real telemetry from NVML, NCCL health-check APIs, or InfiniBand fabric managers. A single NeoCloud node connected to the agent advances this from TRL 4 to TRL 5.
Not a bug — version nomenclature is a product decision, not a technical one. But if a design partner's engineering team runs cargo metadata, they see 0.1.0. Worth aligning with a single source of truth before the first external repository share.
Most deep-tech infrastructure companies raise their Seed round at TRL 4–5 and their Series A at TRL 6–7. TRL 5 with a clean 201-test integration suite, chaos-tested event bus, and a coherent 14-agent architecture is a strong position for first design partner acquisition.
For NeoCloud specifically: what they are evaluating is not whether DeployIQ is production-complete (it isn't — that's what the design partner programme is for) but whether the founding team can get it to TRL 7 on their hardware. The infrastructure TRL 6 evidence answers the "can they ship?" question. The TRL 4 ML and topology layer is exactly where a co-designed design partner relationship adds value — NeoCloud provides real H100 telemetry, DeployIQ closes the gap from TRL 4 to TRL 6 in 90 days.
The ISED Canada TRL framework is also directly relevant to SR&ED (Scientific Research & Experimental Development) tax credits and NRC IRAP funding — both programs fund work between TRL 4 and TRL 7. The current position qualifies DeployIQ for both.