The MILHUB Process
From Forecast to Open Competition
Five steps — see the issue coming, fix the data, prove the part, open the buy. Air Force decisions at every gate. Government-owned data throughout.
Why rebuilding tech data kept failing
- Every part was a bespoke project — manual drawing interpretation made the rebuild cost more than the part was worth.
- Companies bid blind on package completeness, and technical questions (339s) took months to cross organizational lines — firms burned capital waiting, then stopped bidding.
- When reverse engineering did succeed, the data rights often left with the contractor — recreating the sole source it was meant to fix.
What's different: engineer-governed AI agents
- BOM360 already runs in the TITAN environment — every part indexed to its drawings and schematics. The extraction problem is proven, not proposed.
- An agentic workflow does the toil: evaluation agents read, cross-reference, and audit each package against the standard — flagging gaps with confidence scores and drawing-linked evidence.
- Agents draft and flag; engineers decide and sign within their delegated portfolios. Nothing replaces engineering authority — the gain is fewer engineering support requests and fewer hours lost to document archaeology.
- The Air Force rebuilds and keeps its own technical baseline for the usable life of the asset — permanently breaking vendor lock.
The Five Steps
See It. Score It. Fix It. Prove It. Buy It.
Each step produces evidence the next one depends on. The Air Force holds the decision at every gate, and every artifact the process creates stays Government-owned.
Forecast / Identify
Predict part issues before the buy fails. Forecast agents watch demand history, solicitation outcomes, DMSMS cases, and obsolescence signals.
Assess TDP Readiness
Evaluation agents score the technical data — what exists, what is missing, how hard reproduction will be. Built on the BOM360 part-and-drawing index.
Close the Gap
Agents prepare the gap evidence; the cognizant engineering authority approves the rebuild. MILHUB supplements where capability gaps exist. New data stays Government-owned.
Prototype
Air Force-led: prove the part organically through REACT, REARM, and ALC shops. MILHUB adds capacity and advanced-manufacturing depth only as needed.
Source
An Air Force decision: compete among hub manufacturers, or put it on the street for bid. The data package supports either path.
An idea to put on the table: TDP Readiness Levels
TRLs measure technology. MRLs measure manufacturing. Nothing measures whether the data can support a buy.
TDP-RL 1
Absent
No usable technical data — full reverse engineering required
TDP-RL 2
Fragmentary
Major gaps or conflicts — substantial rebuild
TDP-RL 3
Complete, unvalidated
Full definition on hand (2D or model-based) — accuracy unverified
TDP-RL 4
Validated
Verified against configuration — ready to solicit
TDP-RL 5
Production-proven
Has supported build-to-print manufacture and acceptance
Rights ride as a suffix, not a rung
U = Government unlimited rights. L = limited, restricted, or undetermined — releasability determined separately.
AI screens completeness; the engineering authority validates levels 4 and 5. A score is a starting point for a human decision, never a substitute for one.
The Data Work · Ontology and Operating Model
The BOM Is the Spine
An ontology is simply the agreed definition of every object — weapon system, part, drawing, schematic, specification, supplier — and every named link between them. Steps 02 and 03 attach the rest of the technical data to that spine.
Running now
Step 01 — The BOM Spine
Bills of materials generated and delivered across the weapon system portfolio.
Weapon System
MDS / TMS
End Item
Tail or serial number
Assembly
Subsystem and next higher assembly
Part
NSN · part number · CAGE
Bills of materials extracted from Technical Orders are traceable to the figure and callout they came from. This index already exists — it is what the rest of the data hangs on.
Steps 02 and 03
Assess TDP Readiness · Close the Gap
Pull what the standard requires from the systems that already hold it. Then score the rest.
Technical Orders
Figures, sheets, and hotspots — ETIMS, JCALS
depicts → part
Engineering Drawings
2D and 3D and associated lists — JEDMICS, AF-PLM
depicts → part
Schematics
Components and nets, vector-extracted from the sheet
depicts → part
Specs & Standards
MIL-DTL and QPL, material and process specs
governedBy → part
Catalog & Configuration
FLIS and FEDLOG, D200 and D043 I&S, technical baselines
inBaseline → part
Supply & Risk
DLA 339 and ESA dispositions, DMSMS cases, award history
suppliedBy → part
Technical Data Package — scored against MIL-STD-31000C
TDP-RL 1 through 5 per part — what exists, what conflicts, how hard the rebuild really is. A named gap list, each gap routed to the system or authority that can close it.
Complementary by design — four organizations, one model
DSAF
Owns the platform and the graph. Forecasting and readiness analytics run on structured, linked data instead of PDFs.
Engineering
Holds the authority. Agents draft and flag against the standard; engineers disposition and sign.
SASPO
Receives a package that is already complete — DMSMS, analysis and resolution, reverse engineering, and source development.
Procurement
Competes what the data can finally support. Fewer 339s, fewer no-bids, fewer sole sources by default.
Fund the data work once. The bill of materials already reaches the portfolio; steps 02 and 03 turn it into a technical data package the Air Force owns — and every prototype, every buy, and every dollar the hub invests inherits it.
The Workbench
One App That Runs the Process — and the Wing Owns It
The phases of the process run inside one application, deployed in the TITAN environment at IL-5 on NIPRNet, beside the systems of record it reads. Analysts operate it side-by-side from day one.
Watchboard
Forecast agents watch demand history, solicitation outcomes, and DMSMS signals. A part crossing threshold opens a case file in a risk-ranked queue.
TDP-RL Scorecard
Evaluation agents score readiness against the standard; wing engineers validate. Every finding links to its source drawing and revision — checkable in seconds.
Data Room + Gap Worklist
Every drawing, spec, and extraction indexed. Agents route gaps as assigned tasks; CAD and simulation artifacts from the digital engineering toolchain attach to the case.
Build & Qual Tracker
Additive candidacy scorecards, prototype status, and the qualification evidence trail — from first article to acceptance.
Sourcing Package
Controlled export of a review-ready package — recommended input to ESA and DLA AMSC review, handoff to source development. The buy decision stays with the Air Force.
One case file per part
Agents assemble the evidence — scorecards, documents, extractions, analysis, simulation results. People make every decision. One auditable record from risk signal to award, built on the BOM360 index and the problem-parts evidence graph.
No vendor lock, by construction
Deliverables and data rights are defined at award. The wing owns its data, its records, and unlimited rights to every package produced. The result is a working system and a trained team, not a report.
Built to scale across DoD
One case-file standard works for any service's parts — aligned with the hub agenda of bringing problem parts together across the department. The wing owns the reference instance and sets the standard others adopt.
Where the data is ready, industry answers.
Where it isn't, buys fail and sources go sole. The process works that barrier side-by-side with wing teams — predict the part, complete the data, open the competition.