Industrial Edge Intelligence Platform

VANTEDGE Signal. Compressed.
Intelligence. Delivered.

A patent-pending four-stage signal processing pipeline that delivers 32-bit quality intelligence from 8-bit edge hardware — at 80% reduced bandwidth, with deterministic sub-LSB fidelity. No hardware changes required.

See How It Works Request a Demo
80%
Bandwidth Reduction
10B→2B
Per-Sample Payload
8→32
Bit Intelligence Bridge
0
Additional Hardware
The Problem

Industrial sensors are
choking your network

Modern industrial environments run legacy 8-bit hardware that predates today's IIoT platforms — devices that speak Modbus RTU over serial, with no network stack, no floating-point unit, and no path to modern analytics. The standard solutions — hardware replacement, protocol gateways, cloud offload — all require significant investment or infrastructure change.

VantEdge eliminates that requirement entirely. A software-only pipeline that makes legacy hardware a first-class citizen in modern industrial AI pipelines — with no hardware changes, no data loss, and mathematically bounded reconstruction fidelity.

01 / BANDWIDTH
Legacy hardware can't speak modern protocols
8-bit devices output Modbus RTU at 9600 baud — 10 bytes per frame, no OPC-UA, no MQTT, no path to Litmus Edge or any modern IIoT platform without a software bridge.
02 / NOISE
Sensor noise is managed in the wrong place
Filtering and noise removal is typically done downstream — after noisy data has already consumed bandwidth and storage. The problem should be solved at the source.
03 / HARDWARE
Hardware replacement is the assumed solution
The industry assumes legacy hardware must be replaced to participate in modern IIoT pipelines. That assumption is wrong — and VantEdge proves it with a software-only deployment that costs nothing in hardware.
The VantEdge Pipeline

Four stages.
One unified invention.

Each stage transforms the signal and passes its output to the next. The four stages are sequentially interdependent — producing emergent capabilities not present in any individual stage in isolation.

01
RLS Filter
Noise Removal
Recursive Least Squares adaptive filter removes sensor noise while preserving true signal dynamics
02
Horner Model
Derivative + Phase
Polynomial fit + O(N) Horner derivative produces virtual torque and mechanical cycle phase
03
RoPE Encoder
Temporal Encoding
Rotary Position Embedding encodes temporal position within the mechanical cycle — no timestamp needed
04
BitAug-RLT
Compression
Nearest-neighbor lookup in a pre-shared table compresses the embedding to a 2-byte payload
Stage 01 · RLS Adaptive Filter
Recursive Least Squares Refinement
Unlike fixed-window averaging, the RLS filter uses a Kalman-style covariance update with an exponential forgetting factor (λ = 0.97) to continuously adapt to non-stationary industrial signals. No step discontinuities. No fixed-delay artifacts. Just clean signal.
λ = 0.97 · Order 4 · Warm-up aware
Stage 02 · Horner Derivative Model
Virtual Torque — No Sensor Required
A polynomial is fit to a sliding window of filtered samples. Horner's method evaluates the first derivative at the most recent point in O(N) multiplications — producing a real-time torque estimate that eliminates the need for a dedicated torque transducer entirely.
Window = 16 · Degree 3 · O(N) evaluation
Stage 03 · RoPE Encoder
Timestamp-Free Temporal Encoding
Rotary Position Embedding — adapted from transformer language models and applied here for the first time to mechanical cycle phase encoding — rotates signal dimensions by frequency-dependent angles. The dot product of any two embeddings encodes their relative temporal separation without transmitting explicit timestamps.
Dim = 4 · Base = 10,000 · Phase-relative
Stage 04 · BitAug-RLT Compression
2-Byte Payload. Built-In Anomaly Detection.
A pre-shared Remote Lookup Table (built offline via K-Means clustering of representative embeddings) enables nearest-neighbor quantization to a 1-byte index + 1-byte residual delta. Reconstruction error is strictly bounded within deterministic bin precision — below the native noise floor of standard industrial sensors. The delta channel is also your anomaly detector — a large delta means the signal has deviated from the learned model.
uint8 index + int8 delta · 80% bandwidth reduction vs Modbus RTU
Technical Specifications

The numbers
speak for themselves

Bandwidth reduction 80% vs Modbus RTU (10B→2B)
Input protocol Modbus RTU · float32 · serial
Output format 2 bytes (uint8 + int8)
Fidelity model Sub-LSB Deterministic · hash-verified
Reconstruction error Bounded below sensor noise floor
Virtual torque range 40–55 N·m (no sensor)
RLS forgetting factor λ = 0.97
Platform compatibility 8-bit MCU · MQTT · Docker
Integration Litmus Edge · Databricks
IP status Patent Pending
Bandwidth Comparison · Per Sample
Standard Modbus RTU frame 10 bytes
VantEdge compressed payload 2 bytes
# VantEdge pipeline output — per sample stage_1: RLS refined signal → float32 stage_2: virtual torque → 45.3 N·m stage_3: RoPE embedding → [0.82, -0.31, 0.55, 0.17] stage_4: compressed payload → [idx: 0xA4] [Δ: +3] TRANSMITTED: 2 bytes ← was 10 bytes (Modbus RTU) ANOMALY SCORE: 0.12 ← nominal
Applications

Built for the
industrial edge

⚙️
Rotating Asset Monitoring
Bearings, fans, pumps, and motors. VantEdge extracts virtual torque from vibration data and detects developing faults through the anomaly channel — before failure occurs.
Predictive Maintenance
📡
Constrained Wireless Networks
Wireless sensor nodes with limited radio bandwidth or battery budgets. VantEdge's 80% payload reduction vs standard Modbus RTU directly extends sensor network range and battery life — with no loss of signal intelligence.
IIoT Infrastructure
🔋
Battery Cell Manufacturing
High-density sensor arrays in EV battery production lines. Real-time signal quality at reduced data center ingestion cost — proven in one of the world's most demanding manufacturing environments.
EV / Energy
Bill Edenfield

IMG_0055.jpeg

Founder
Bill Edenfield
Founder & CTO · VantEdge Intelligence

Bill Edenfield spent years managing and overseeing production systems where the quality and latency of sensor data had direct consequences on output quality and throughput.

The persistent challenge of managing noisy, bandwidth-heavy sensor streams in constrained industrial environments became the foundation of VantEdge Intelligence. Bill designed and built the entire four-stage pipeline — from mathematical concept to working proof-of-concept — delivering deterministic sub-LSB fidelity and actionable signal preservation from legacy hardware that the industry assumed required replacement. He is currently validating the technology with leading industrial IoT platform partners.

VantEdge Intelligence is based in the Atlanta Metropolitan Area and is actively seeking partnerships with industrial automation OEMs, IIoT platform providers, and edge hardware manufacturers.

Founder & CTO Patent Pending Atlanta, GA IIoT · Edge AI
Get In Touch

Ready to see
VANTEDGE in action?

Whether you're an industrial automation OEM, an IIoT platform provider, or an engineering leader exploring edge optimization — we'd love to show you a live demonstration of the pipeline.

Or connect directly on LinkedIn  ·  VantEdge Company Page