Belleau Labs

Methodology

Two scoring rubrics. Five dimensions each. Twenty-five points max.

Govini scores contractor-fit — does your company match this RFP? We score AI-readiness — is this RFP actually shaped for an AI-native startup, or dressed-up legacy work? Different question. Different answer. The two rubrics below are how we operationalize that distinction for contracts and for early-stage companies.

Contracts rubric

Applied to every DoD solicitation. Scores 0-5 on each of five dimensions for a maximum of 25. The score is the headline; the rationale paragraph is the value — it explains the tradeoffs behind each number.

21-25
Top-tier
16-20
Strong
11-15
Mixed
0-10
Low signal
Dimension0 — Not present3 — Partial5 — Full signal
Pathway SpeedCost-plus FAR with traditional acquisition pathwaySBIR Phase II.5, Direct-to-Phase IIOTA, AFWERX STRATFI/TACFI, DIU CSO, OSC, Replicator
Timeline RealismImpossible or absurd deadline14-30 day response, tight but achievable45+ day response window AND prototype timeline that respects AI dev cycles
Problem Framing"AI/ML" used as marketing, no real problem statedReasonable framing but scope ambiguityBounded problem + named end-user + success metrics + data described
AI / ML Fit"AI" is marketing; problem is not actually AI-shapedAI is meaningful but secondary to integration / pipeline workCore problem is genuinely AI-shaped AND data is named/available
Award + TransitionStudy contract with no follow-on path$1M-$5M ceiling OR vague transition mention$5M+ ceiling AND explicit production transition pathway

Example record

CRAZY TRAIN — Umbra Lab
AFWERX STRATFI · $60M ceiling · Sep 2025
20/25
Pathway Speed
5
Timeline Realism
3
Problem Framing
4
AI / ML Fit
3
Award + Transition
5

AFWERX STRATFI is one of the premier fast-track OT instruments available, earning a top pathway score, and the $60M ceiling with an explicit TRL 7 on-orbit prototype culminating in a 6-month operational demonstration represents a clear, substantial transition pathway. AI/ML fit scores a 3 because wh…

Seed rubric

Applied to early-stage defense tech companies. Scores 0-5 on five dimensions that predict which seed-stage companies will reach $10M ARR or a government prime contract within three years.

20-25
Track closely
15-19
Monitor
10-14
Early signal
5-9
Low priority
Dimension0 — Not present3 — Partial5 — Full signal
Team StrengthNo founders named, generic descriptionCapable team, unclear domain fitNamed founders with deep defense or domain expertise (ex-military, prior exits, domain PhD)
Validation DepthNo evidence of tractionOne form of validation (SBIR Phase I, small seed round)Government contract + external VC funding + named pilot customer
NoveltyCommodity offering, no visible differentiationSolid product in a crowded spaceClear technical differentiation in underserved defense niche
Market TimingNo clear path to DoD budgetAddresses real defense need, no recent pull signalDirectly addresses a named DoD priority or SECDEF initiative (last 18 months)
Capital EfficiencyUnclear financials or excessive early dilutionReasonable capital deploymentMeaningful product shipped on under $3M raised

Example record

Red Six Aerospace
SBIR Phase I · Seed · Aug 2022
22/25
Team Strength
4
Validation Depth
5
Novelty
4
Market Timing
5
Capital Efficiency
4

Red 6 shows exceptional validation depth with multiple Phase 2 SBIR awards (TACFI, STRATFI) and direct integration into AFSOC MC-130J platforms, plus recent 2024 momentum on ATARS. Team strength appears solid given Phase 2 progression and AFSOC/AETC partnership, though founder names and backgrounds …

Data sources

SourceWhat it coversUpdate cadence
SAM.govFederal opportunities — OTA, CSO, BAA, SBIR, sources soughtDaily
DIUCommercial Solutions Opening awards and project pagesWeekly (press releases)
AFWERXSTRATFI ($3M-$15M), TACFI ($375K-$2M), Spark awardsWeekly (press releases)
OSCOffice of Strategic Capital direct DoD loans and investmentsAs published
DARPABAAs, program announcements, performer newsAs published
Y CombinatorYC-backed defense tech companies from batch announcementsPer batch (bi-annual)
SBIR Phase IEarly-stage awardees — DoD and DoE programsPer award cycle
TechCrunchFunding rounds for defense-adjacent tech companiesAs published
DefenseScoopProgram awards and defense tech funding newsAs published

Update cadence

Scrapers run nightly. Scoring runs weekly — each new batch goes through the rubric and the LLM rationale. A manual review pass covers the top-scored records before they go into the newsletter: if a record scores 18+ but the rationale reads off, it gets flagged and re-scored. The database compounds over time; old records stay in place even as new ones are added.

Limitations

LLM scoring is calibrated against domain intuition, not against actual transition outcomes. A contract that scores 22/25 might not attract a single qualified bidder if the program office has already pre-selected a vendor. A company that scores 18/25 might still run out of runway.

The rubric will be recalibrated as score-to-outcome data accumulates over six to twelve months. Dimensions that don't predict outcomes will be dropped or reweighted. v0.1 scores are a starting point, not a final answer.

Some data sources rate-limit or publish inconsistently — ingest can be partial. SAM.gov in particular returns incomplete descriptions that require a separate fetch. If a record looks thin, check the source link on the detail page.