Greenlight Industries · Seed round · Investor walkthrough

Star Trek replicator —
minus the sci‑fi.

On‑demand manufacturing of engineered composite structures, from tape. CAD in, finished part out. Zero touch‑labor. Sealed environment. On‑site or remote‑operated. A NASA‑spinoff stack of hardware, telemetry, and AI that gets smarter every time it builds.

We are designing a machine to give the AI all the tools it needs to make and manufacture the next generation of structures. — Dr. Robert Bryant · 2026‑05‑24. Lead with the AI; the hardware is the substrate that makes it possible.

A 6‑minute walkthrough. Six weeks of a composites engineer's life — compressed into a single afternoon by a TapeLayer™ in a shipping container.

Start the walkthrough → Skip to the ask
Pillar 1 · Digital Twin

A virtual avatar of each physical machine. Simulates the build before any material is loaded. Captures telemetry during the build. Freezes the parameter set as a reproducible recipe on QC pass.

Pillar 2 · Experimental DB

"As‑made" property tables from real specimens, tested to ASTM, MIL‑HDBK, and ISO. Open data in. Subscription data out. Only accepts data from Greenlight machines.

Pillar 3 · Machine AI

CAD in, layup plan out. Customer uploads geometry and targets — cost, speed, weight, strength, safety margin. AI selects the optimal layup, single material or hybrid, and the twin builds it.

01 · Cold open

Sarah is choosing between three carbon‑fiber prepregs.

She's a composites engineer at an aerospace OEM. The part is a wing skin. She has three vendor datasheets, two NDAs, a spreadsheet, and six weeks to make a call.

Different test conditions on every datasheet. Different temperatures, different specimen geometries, different standards. No way to compare apples to apples without redoing the coupon tests herself. So she does. And the calendar slips.

$8T/yr

Wasted in global manufacturing

Forbes / Mingo industry estimate. Most of it is rework, scrap, and over‑built parts.

$2M+

Legacy ATL/AFP machine

One part, one material, one facility. No telemetry portability. No hybrid materials mid‑build.

80%+

Composite build time is touch‑labor

Hand layup is still the dominant process. Quality varies with the technician.

"With this data and processing conditions, we can compare it with what we get from our machine. This will lead to quantifiable cost savings even for simple test specimens." — Dr. Robert Bryant, Founder/CEO · 2026‑05‑15 evaluation
02 · Greenlight enters

The TapeLayer™ — a fab in a shipping container.

Self‑contained. Environmentally‑controlled. Hybrid materials mid‑build. Zero touch‑labor. CAD in, finished engineered part out. Drop it on a flight line, a factory floor, or a forward operating base.

TAPELAYER · TEU CAD → PART TELEMETRY UPLINK Schematic — full TapeLayer™ reel: Biz Plan/TapeLayer Video.mp4 ▷ Reel · 80 MB

Two product configurations: Compact (desktop, ~$40K) and Standard (TEU‑scale shipping container, up to $3M). Same software stack. Same telemetry contract. Same digital twin.

Not a 3D printer. The TapeLayer™ is a coordinate‑axis tape‑laying machine — it lays continuous engineered fiber and can insert contoured tool surfaces mid‑build. It places structural material along load paths, not melted plastic layer by layer.

03 · SpecSniper — Pillar 2 in action

Sarah types "carbon fiber epoxy prepreg."

Six weeks of vendor calls collapses into four seconds. SpecSniper reads the real datasheets, extracts ASTM‑standardized test rows, and writes them into one normalized table with every value provenance‑linked to the source PDF, page, and verbatim quote.

SpecSniper · datasheet-ai-c24ku24zea-uc.a.run.app LIVE
5

Curated vendor PDFs

Company A, Company B, Company C, plus reference standards. (Vendor names obfuscated for confidentiality.)

462

Standardized rows extracted

Every row keyed to standard, temperature, conditioning, and source quote.

~4s

Cold‑load time

Pre‑warmed sample cache baked into the container image (v3.4, 2026‑05‑17).

"This is exactly what I wanted… real time saver." — Dr. Robert Bryant · 2026‑05‑15 evaluation of Demo E

Open the live tool in a new tab: SpecSniper →

04 · Strata — Pillar 2/3 bridge

Filter. Compare. Drill back to source.

Spreadsheet ergonomics on the front end. Provenance on every cell. Subscription gate around the data egress. Strata is the same table Sarah filters and the table the Machine AI optimizes against in Pillar 3 — and it's a sellable product in its own right, parallel to the AI.

Strata · db-explorer-c24ku24zea-uc.a.run.app LIVE
"Deliminated so spreadsheets can digest it." — Dr. Robert Bryant · 2026‑05‑15 call

Open the live tool in a new tab: Strata →. Only accepts data from Greenlight machines — open data in, subscription gate on data out.

05 · Optimax — Pillar 3 precursor

Build it on our machine. Then tune it.

Sarah loads the target spec, runs a batch on the TapeLayer™, and Optimax matches the run to the target row‑for‑row. Where the machine deviates, Claude proposes parameter adjustments — with explicit ply layup — and estimates the per‑batch cost savings.

Optimax · machine-optimizer-c24ku24zea-uc.a.run.app LIVE
In spec
23 / 27

Tests within the combined‑σ envelope. Target: Company A CF/epoxy prepreg (grade A‑1). (vendor obfuscated)

Borderline / Out
3 + 1

All trace to an under‑cure + over‑pressure story. Five designed deviations in the synthetic run.

Recommendation

+10 °C cure · +30 min dwell · −1 bar pressure

$7,560 per‑batch savings · −25% time‑to‑spec (4 → 3 iterations).

"With this data and processing conditions, we can compare it with what we get from our machine. This will lead to quantifiable cost savings even for simple test specimens." — Dr. Robert Bryant · 2026‑05‑15

Open the live tool in a new tab: Optimax →. Excel companion pack — five sheets, live formulas — at /Greenlight/machine-optimizer/output/machine_optimizer_synthetic_data_2026-05-15.xlsx.

06 · Avatar — Pillar 1 in action

Predict the part before you build it.

Avatar is a virtual version of each physical machine — straight out of Robert's car‑telemetry analogy. Load the geometry. Color‑map the as‑made material properties. Run a synthetic load case. Watch failure modes surface — with error bars, not certainty theater.

Avatar · digital twin viewer — LIVE · best opened full‑screen
σ_max · 92% UTS ±10% · ASTM D3039 low stress nominal approaching limit hotspot Wing skin · Company A CF/epoxy · 8‑ply quasi‑isotropic · 14 kN tip load

Schematic of the live viewer. Avatar surfaces failure modes with explicit error bars (±10% when a vendor std dev is absent), recommends a recipe from Pillar 3, and roundtrips back into Pillar 1 simulation before any tape is loaded.

"We're developing a digital twin that is reliable by any standard." — Dr. Robert Bryant · 2026‑05‑15

Open the live viewer in a new tab: Avatar → (3D viewer renders best in its own window).

07 · The moat — Bryant's analogy

A car with telemetry knows itself. Another car can't use that data.

From Robert's April 3 2026 email — the foundational framing for the Greenlight moat. The same logic that makes a sensor‑instrumented car better at predicting its own routes than Google Maps makes a Greenlight TapeLayer™ better at predicting its own parts than any generic model.

Frame 1

A car learns its routes

Onboard sensors capture speed, drag, payload, terrain. The car builds a model of itself, its driver, its load.

Frame 2

A virtual avatar emerges

Trip predictions tuned to this vehicle — beating generic mapping software that doesn't know the car.

Frame 3

The data can't transfer

Plug the dataset into another car: it doesn't fit. Different sensors, different actuators, different physics.

Frame 4

Greenlight: same logic

Each TapeLayer™ tunes its own twin. Datasets compound per‑machine. AI gets better only on Greenlight hardware.

M1
TapeLayer #1
tl-001 · 2026‑Q3

412 builds. 18 recipes locked. Twin agreement vs. coupon: 96.4%.

M2
TapeLayer #2
tl-002 · 2026‑Q4

218 builds. 11 recipes locked. Twin agreement vs. coupon: 94.1%.

M3
TapeLayer #3
tl-003 · 2027‑Q1

67 builds. 4 recipes locked. Twin agreement vs. coupon: 91.7% — still climbing.

Material‑agnostic · NDA

One machine, any material. Competitors lock to one.

Electro Impact and Coriolis build material‑specific machines. The TapeLayer™ is material‑agnostic — which is exactly why a customer can't fork our AI onto someone else's hardware. The AI is trained on what our agnostic machine can do; material‑specific rigs can't reproduce it.

Coupled by design

The software doesn't run standalone.

Greenlight software is coupled to a specific TapeLayer™ — not sold on its own. That's the answer to "why doesn't your software run on other machines?": the per‑machine twin can't transfer because it's tuned to its own sensors and actuators.

The car learns your routes, builds a virtual avatar, and predicts trips better than Google Maps — because it's tuned to your vehicle. Someone else's car can't reuse the data. It doesn't fit their car. — Dr. Robert Bryant · 2026‑04‑03, paraphrased from the "Software slides discussion" thread
I've got the data that the insurance company needs to ensure my product. — Dr. Robert Bryant · 2026‑05‑15. Right‑sized parts with calibrated uncertainty are insurable in a way over‑built guesswork is not.

Audience rule — follow the money. A workable twin reduces material consumption and rework. The losers are materials suppliers who today benefit from over‑built parts. Greenlight pitches OEMs, manufacturers (we are our own first customer), end‑users, funders, and insurance companies — never suppliers. — Robert, 2026‑05‑15.

08 · Economics

Three streams. Hardware sells the relationship; software is the annuity.

Stream 1 Direct sales

TapeLayer™ Compact (desktop, ~$40K) and Standard (TEU‑scale, up to $3M). Hardware margin, service contracts, consumables.

Stream 2 Manufacturing‑as‑a‑Service

Lease or rental at the customer site, or remote‑operated. Customers send CAD; we ship parts. Recurring revenue, higher lifetime value.

Stream 3 SaaS subscription

Gates the AI, the aggregated database, the optimizer. Greenlight‑machine‑only data is the moat. This is the annuity.

Network effect

More machines More recipes locked Better Machine AI More parts shipped More machine sales
09 · Same OEM. Six weeks later.

Sarah has shipped.

She picked a material in an afternoon, not six weeks. Her run matched target on 23 of 27 tests on the first batch. The optimizer flagged the three borderline cases, recommended +10 °C and +30 min, and dropped per‑batch cost by $7,560. The twin signed off on the wing skin. The part flew.

"This is exactly what I wanted… real time saver." — Dr. Robert Bryant · 2026‑05‑15
10 · The ask

$6M seed at $30M post · 20%

Phase 0 ($200K, self‑funded) is complete. Seed funds Phase 1: prototype Compact + Standard builds, beta program, Andrews‑Cooper engineering, tooling, process optimization.

Raise
$6.0M
Post
$30.0M
Allocation
75% R&D
Operating
25%

Delaware C‑Corp · Virginia HQ · Founder/CEO: Dr. Robert Bryant (30+ NASA Langley years; National Inventors Hall of Fame; sole inventor on two NASA patents on the TapeLayer™). CFO/IR: David B. Smith. Co‑founder/Legal: Richard R. Newton. Build partner: Andrews‑Cooper.

Request the deep‑dive → gl‑ind.com
step · Esc top