The DeepSnow platform: how a wet lab plus an AI polymer discovery engine produces a real product roadmap

TL;DR. DeepSnow is a polymer discovery platform with three integrated layers: an in-house wet lab, an AI polymer discovery engine, and a product portfolio. The wet lab synthesizes and tests candidate polymers. The discovery engine ranks candidates and proposes the next experiment, IP-aware. The product portfolio is what ships. SL6733 is product 1 (snowmaking additive, EU lab pilots 2026/27). DS-100 series sAFGPs and DS-400 (ice rink IRI) are product 2 and 3, designed by the engine. This article describes how the layers fit together and why this composition compounds.

The pitch in one paragraph

DeepSnow is a vertically-integrated polymer discovery platform. We operate a wet lab (chemistry, synthesis, characterization), an AI discovery engine (software that ranks and proposes polymer candidates), and a developing product portfolio (snowmaking, ice rink, cold chain). The three layers run in a closed loop: the engine proposes, the lab tests, the lab data refines the engine, and validated candidates become products. The first commercial output is SL6733; the second is the DS-100 series; the third is DS-400; the platform's value lies in the production rate beyond those.

Why three layers, not one

A lot of materials-discovery efforts try to do one of these three things in isolation. The pattern that fails:

Wet lab only: domain expertise, can build products, but cannot systematically explore design space. Output is bounded by what individual chemists can intuit.
AI only: can propose vast candidate libraries, but without wet-lab feedback the predictions degrade. The model trains on stale assumptions.
Product company only: ships what works today, but no ongoing pipeline. The next product is contingent on someone else's research.

The pattern that compounds is all three integrated. The engine proposes more candidates than any human team could systematically test. The lab generates the data the engine needs to keep its predictions calibrated. The product portfolio is what monetizes the platform — and provides the operational discipline that keeps R&D from drifting into purely academic exploration.

The wet lab layer

DeepSnow's wet lab runs polymer synthesis, characterization, and IRI/freeze-thaw testing. The standard capabilities:

Free-radical polymerization for ultra-high-MW polyacrylamide synthesis. The SL6733 chemistry — anionic poly(acrylamide-co-sodium acrylate) at 15–20 MDa with 30–40 mol% sodium acrylate — is produced here at pilot scale.
NCA polymerization for synthetic polypeptide synthesis. The DS-100 series sAFGPs (Ala/Glu alternating backbone) come from this line.
AF4-MALS for absolute molecular-weight characterization. The gold standard for ultra-high-MW polymers, where standard GPC fails.
Splat assay infrastructure for IRI quantification. Polarized-light microscopy on temperature-controlled stages.
Freeze-thaw + ice nucleation testing for product-relevant performance characterization.

Capacity: pilot batches up to 100 L for resort-trial volumes.

Everything that goes into the discovery engine's training data comes from this lab. Everything that comes out of the discovery engine gets tested here. The two are tightly coupled.

The discovery engine layer

The engine is software. It does three things:

Candidate proposal: given a target property profile (e.g., "synthetic antifreeze polypeptide with maximal IRI activity, manufacturable via NCA, engineered around IP X, Y, Z"), generate ranked candidate sequences or architectures.
Property prediction: for each candidate, predict ice-binding affinity, IRI potency at target concentration, manufacturability, and solubility. Models combine sequence-to-property predictions (machine learning) with thermodynamic / molecular-dynamics constraints.
Experimental design: choose the next experiment to run in the wet lab, optimizing for information gain — i.e., learn the most about the design space per synthesis run.

The engine is IP-aware by design. Existing patents in the antifreeze-protein and IRI-polymer space define exclusion zones in sequence/composition space. Candidates that fall inside those zones are flagged. Candidates that fall outside — but still capture the underlying mechanism — are prioritized. This is what makes the discovery engine commercially actionable rather than purely academic.

The product portfolio layer

A real product portfolio is what differentiates a platform from a research effort. DeepSnow's portfolio:

SL6733 — snowmaking polymer additive. The lead product. Polyacrylamide-co-acrylate + cold-water-swelling starch. EU lab pilots targeted 2026/27 season; commercial deployment 2027/28. Engineered via wet-lab chemistry; the engine's role here is incremental optimization (charge density, MW distribution, residual monomer specs).
DS-100 series — next-generation sAFGP. Polypeptide chemistry. R&D stage. The engine is what designs the specific sequences. 91–94% MGS reduction at 100 µg/mL in lab assays. Pilot timeline 2027/28+.
DS-400 — ice rink IRI. Surface-grafted IRI polymer for ice rink resurfacing systems. R&D stage. Different application surface (rink ice vs falling snow) with different operational profile (surface-applied vs water-injected).
Beyond — adjacent verticals from the same platform: food-IRI (cold-chain food handling), cryopreservation (biologics cold-chain), surface IRI (cold-chain logistics).

The product portfolio is what gives the discovery engine a north star. Without that, R&D drifts.

What compounds

Three things, in order:

Each product feeds the next. The wet-lab knowledge from SL6733 production refines the engine's predictions for DS-100. The DS-100 lab data refines predictions for DS-400. The accumulation of cross-product chemistry data is what makes the platform durable.
The IP-aware design surface gets wider over time. Every patent landscape DeepSnow maps for one product becomes an input for the next. The engine accumulates structural knowledge of what is and is not defensible.
The product portfolio diversifies revenue. SL6733 has snowmaking-industry exposure. DS-400 has ice-rink-industry exposure. Food-IRI has cold-chain exposure. The same chemistry stack across multiple end-markets is a substantially more resilient business than a single-product play.

What this is not

A platform claim is a strong claim and we want to be honest about what we have not yet shipped:

The discovery engine is not a black box that magically produces optimal candidates. It is software with measurable error rates, and it gets better with more wet-lab data. We are not the AlphaFold of polymers — we are a domain-specific tool that does well in a narrow but commercially valuable region of design space (IRI-active polymers and polypeptides).
SL6733 is not shipping yet. It is in active R&D with EU lab-pilot trials targeted for 2026/27. The commercial timeline is 2027/28.
DS-100 and DS-400 are at earlier R&D stages. Lab-pilot data is in hand; pilot trials with operators are 2027/28+.

The platform compounds, but the compounding takes seasons.

Where to engage

If you are a resort operator, polymer chemist, ice scientist, or investor curious about how the platform works — use the contact form. For pilots, the cohort signup is here.