// hardcoded — pre-seed pitch
On-Device NLP SDK

AI that runs
on the device.
Not in the cloud.

A lightweight SDK that gives any mobile or web app
natural language understanding — fully offline,
no API keys, no per-query costs, no data leaving the device.

Live demo: demo-wallet.lightsoftware.tech
HARDCODED
01 / 11
// the problem

Every app that wants
conversational AI
hits the same wall.

Developers have two options today — both are broken.

Cloud AI APIs
$0.01–$0.10 per query · user data leaves device · breaks offline · compliance nightmare for regulated apps
Build it in-house
Requires ML team · 6–12 months · ongoing maintenance · most teams simply can't do it
Do nothing
App stays menu-driven while competitors add conversational UX. Users leave.
02 / 11
// the solution

Hardcoded removes the tradeoff.

→
100% on-device processing. User data never leaves the phone. Ever.
→
No backend required. No API keys, no cloud infrastructure, no ops overhead.
→
Flat licensing. No per-query fees. 100K daily active users costs the same as 100.
→
Works fully offline. Travel apps, field apps, areas with poor connectivity.
→
Days to integrate. Swift package / npm / AAR. Call sdk.process(userText) — get structured JSON back.
→
Multi-turn dialogue. SDK asks follow-up questions to collect missing entities. No server round-trips.
03 / 11
// technology

The hard part:
≤10MB.
<200ms.

Standard transformers (BERT, DistilBERT) run 60–400MB. Getting to 10MB without accuracy collapse requires solving three problems simultaneously.

RUNTIME
Rust → WASM — single binary, all platforms
↕
MODEL
INT4/INT8 quantization + knowledge distillation
↕
DIALOGUE
Stateful multi-turn — in-process, no server
↕
PLATFORMS
iOS (CoreML) · Android (NNAPI) · Web (WASM)
04 / 11
// how it works

Three steps.
Days to integrate.

1
Add the SDK
Swift Package / npm package / Android AAR
2
Define your intents
JSON config — the commands your app supports and what parameters each needs
3
Call one function
// TypeScript example
const result = sdk.process("send 500 USDC to Pete");
// result.intent → "transfer_funds"
// result.entities → {amount: 500, currency: "USDC", recipient: "Pete"}
// result.status → "Done"
05 / 11
// market

The market is
any app that can't afford
to trust the cloud.

$8B
Global NLP market 2024 — growing 20%+ annually
60M+
Trust Wallet users — one wallet, one vertical
$0
Per-query cost after integration. Forever.

Primary verticals: crypto wallets, fintech, healthcare, enterprise tools — where sending user commands to OpenAI is architecturally or legally unacceptable.

06 / 11
// business model

Licensing, not
metering.

On-device means we can't track usage server-side. Every model is trust-based licensing — like Dolby, ARM, or HDMI.

Per-unit royalty
Dolby / ARM model. Fee per app release or device installation. Scales with the customer's success.
Flat annual license
Per app. Predictable for customers. Replaces $50K+/year in cloud API costs at a fraction of the price.
Open core
Free SDK drives developer adoption. Paid enterprise tier adds custom training, SLAs, compliance docs.

Final model to be determined after pilot phase validation.

07 / 11
// traction

Already
moving.

Pre-product. Pre-revenue. But not pre-validation.

Live web demo
Web3 wallet demo at demo-wallet.lightsoftware.tech
iOS build available via TestFlight
Signed pilot — United Network
Multi-chain NFC hardware wallet on iOS & Android
Zero data collection — perfect architectural fit
10+ named pipeline targets
Exodus, Rainbow Wallet, Phantom, Trust Wallet
Direct founder contacts mapped, outreach underway
Platforms complete
iOS ✓ · Android ✓ · Web/WASM ✓
08 / 11
// team

Two engineers.
25 years of production systems.

Sergei Sukhanov
CEO & Co-founder
25 years in software. Started in military software systems — environments where data isolation is non-negotiable. Built blockchain infrastructure at protocol level: consensus mechanism work in Monero, smart contract systems on TVM-based chains. Leads business development and all customer conversations.
Alexey Romanov
Co-founder, R&D
Head of Analytics at Ibisse Blockchain Labs. Background in Python, infrastructure, DeFi R&D. Core strength: navigating open-ended technical problems from "we don't know if this is possible" to working prototype. Leads model compression research and runtime architecture.
09 / 11
// vision
"The MP3 codec existed in 1993.
The first player appeared in 1995.
By 2000, it had restructured an industry."

On-device AI inference is in its 1993 moment. The hardware is ready. The infrastructure layer doesn't exist yet. The endgame is an SDK that runs on a $30 IoT device anywhere in the world — giving a market trader in Karachi the same voice interface that enterprise software charges thousands for.

NOW
Pilot with United Network. R&D on model compression.
2025–26
3–5 paying wallet & fintech customers. Seed round.
2027+
Horizontal expansion. IoT. Any device, any language.
10 / 11
// the ask
Pre-seed round
$500K
SEEKING · 18-MONTH RUNWAY
60%
R&D — model compression, runtime engineering
25%
Pilot support & customer development
15%
Operations & legal (US entity setup)
Contact
Sergei Sukhanov — CEO
Demo: demo-wallet.lightsoftware.tech
Pilot: United Network (united.network)
Platforms: iOS ✓ Android ✓ Web ✓
→ VIEW LIVE DEMO
11 / 11