Turnkey Device • NHTSA-Aligned • Patent Pending

Breathalyzers Can't Catch What's on the Road Today.

Recreational legalization and polydrug use are rising fast. A driver on alcohol, cannabis, and a prescription benzodiazepine can fall below every legal cutoff — and still be dangerously impaired. Chemical thresholds were never designed for this.

Functional testing measures what actually matters: the ability to drive. Oculometrix is a purpose-built device that brings objective, AI-powered SFST testing to every traffic stop.

Request a Demo Unit See What Ships

4 SFST Tests Built In

60 Hz Real-Time Analysis

0 Setup Required

Oculometrix device — purpose-built phone with tactical holder showing SFST test interface in kiosk mode

What Ships in Every Kit

Oculometrix is a turnkey device, not an app download. Every unit arrives ready to deploy with zero configuration.

Dedicated iPhone

A pre-configured iPhone with TrueDepth and LiDAR cameras, optimized for on-device ML inference. Locked in single-app kiosk mode — officers cannot exit, browse, or alter settings.

Tactical Holder

A ruggedized grip mount designed for roadside conditions. Positions the device at the correct distance and angle for NHTSA-standard eye tests. Protects against drops and weather.

Pre-Installed AI Software

All neural networks, pose-estimation models, and SFST protocols ship pre-loaded and validated. Over-the-air updates keep models current without officer intervention.

Why Chemical Tests Alone Are No Longer Enough

Legal thresholds were designed for single substances. Today's drivers mix alcohol, cannabis, opioids, and prescriptions — falling below every cutoff while dangerously impaired. Functional testing is the only way to catch what chemistry misses.

Arbitrary Cutoffs Fall Short

Legal limits for individual substances — 0.08 BAC for alcohol, 5 ng/mL for THC — were set in isolation. When a driver combines alcohol, cannabis, and prescription medications, each analyte may fall below its legal threshold while the combined effect severely impairs driving performance.

Chemical Tests Miss Polydrug Impairment

Breathalyzers detect only alcohol. Oral fluid tests screen for specific analytes. Neither measures actual impairment. A driver at 0.05 BAC with THC and benzodiazepines may be more dangerous than one at 0.10 BAC alone — yet passes every chemical threshold.

Functional Testing Measures What Matters

Roadside sobriety tests assess the skills driving actually demands: gaze stability, postural sway, gait symmetry, and divided attention. The Oculometrix device applies computer vision and ML to these tests, delivering a substance-agnostic impairment assessment regardless of which drug or combination is involved.

See How the Device Works Request a Demo Unit

How It Works

Three steps from traffic stop to court-ready evidence — using the device that ships to your department.

Power On & Select

Power on the Oculometrix device — a dedicated phone with pre-trained ML models loaded in kiosk mode, secured in the tactical holder. Select the test protocol and the AI pipeline initializes automatically.

Administer Test

The device runs NHTSA-standard protocols while on-device neural networks perform real-time inference. Computer vision models capture gaze vectors at 60 Hz; pose-estimation algorithms track skeletal landmarks for balance and gait analysis — all processed locally with zero cloud latency.

Review Evidence

The ML pipeline outputs a binary PASS/FAIL classification backed by feature-level confidence scores, time-series signal charts, and encrypted telemetry — ready for court proceedings.

Four NHTSA Tests. All Built Into the Device.

Every Oculometrix unit ships with dedicated ML models for each Standardized Field Sobriety Test — pre-loaded and ready to run.

Smooth Pursuit

A gaze-tracking neural network follows the subject's eye movements against a moving stimulus. Signal-processing algorithms extract phase correlation, amplitude ratio, and saccadic intrusion rate — feeding a classifier that detects loss of smooth pursuit.

Computer Vision • Front Camera

HGN Nystagmus

Deep-learning feature extraction identifies involuntary nystagmus during horizontal gaze. The model evaluates all 6 HGN clues per NHTSA standards — lack of smooth pursuit, distinct and sustained nystagmus at maximum deviation, and onset prior to 45° — with sub-pixel precision.

Deep Learning • Front Camera

One-Leg Stand

A pose-estimation model tracks 17+ skeletal keypoints during a 30-second single-leg stance. ML classifiers score sway amplitude, hops, arm raises, and foot drops — achieving 89.5% precision on validation data.

Pose Estimation • Rear Camera

Walk & Turn

Sensor-fusion algorithms combine pose-estimation and accelerometer data to count heel-to-toe steps and classify 8 NHTSA clues during a 9-step tandem walk. The multi-signal ML pipeline achieves 100% precision with 80.6% recall.

Sensor Fusion • Rear Camera

The Oculometrix AI Advantage

Traditional field sobriety tests rely on subjective officer observation. The Oculometrix device replaces guesswork with machine learning — a single piece of equipment that brings objective, quantitative intelligence to every roadside assessment.

Turnkey hardware kit. Phone, tactical holder, and pre-loaded ML models ship in one package. Power on and go — no app store, no cloud dependency, no IT setup.
Court-ready AI output. Encrypted recordings, model confidence scores, feature-level telemetry, and time-series signal charts provide irrefutable, explainable evidence.
Substance-agnostic detection. Computer vision models detect impairment patterns from cannabis, opioids, stimulants, and poly-drug use — substances breathalyzers miss entirely.
ML models validated on peer-reviewed datasets. Detection classifiers benchmarked on PhysioNet and Health & Gait corpora with automated regression suites and cross-validation.
Binary classification. No ambiguous results. Every test produces a definitive PASS/FAIL prediction with an associated confidence score.

6 HGN clues classified
Neural-network feature extraction

8 Walk & Turn clues detected
Sensor-fusion ML pipeline

89.5% OLS classifier precision
Cross-validated on PhysioNet

60 Hz Real-time inference rate
On-device edge AI

Oculometrix vs. Traditional Methods

Capability	Traditional SFST	PBT / Breathalyzer	Oculometrix
Detects alcohol impairment	Subjective	Yes	Yes — objective
Detects drug impairment	DRE only	No	Yes
Objective measurement	No	BAC only	Full ML telemetry
Court-admissible data	Officer testimony	BAC reading	AI-generated charts + encrypted data
AI / Machine learning	None	None	On-device neural networks
Special hardware required	None	Breathalyzer unit	Purpose-built device
Calibration needed	N/A	Regular	None
Per-test consumables	None	Mouthpieces	None

AI in the Field

Representative use cases showing how machine learning transforms roadside assessments.

Polydrug Detection

A suspect blows 0.00 on the PBT but appears visibly impaired. The Oculometrix gaze-tracking model flags anomalous saccadic patterns and loss of smooth pursuit. The subsequent toxicology report confirms fentanyl. AI catches what breathalyzers cannot.

Explainable AI in Court

Instead of relying solely on officer testimony, the arresting officer presents ML-generated signal charts, feature-level confidence scores, and quantitative telemetry. Explainable model outputs make the evidence significantly harder for defense counsel to challenge.

Accelerated Training

Traditional SFST training takes weeks to produce consistent results. With AI-guided protocols built into the device, newly trained officers produce reliable, model-scored assessments from their first deployment — the neural network compensates for human variability.

Cost-Effective Edge AI

Competing solutions require $3,000+ VR headsets with cloud-dependent AI. Each Oculometrix device ships as a self-contained kit at a fraction of the cost, making department-wide deployment of AI-powered testing feasible.

Walk & Turn Validation

An officer manually counts 18 steps during a tandem walk. The sensor-fusion ML pipeline counts 17 — corroborating the officer’s observation — and the pose-estimation classifier flags arm-raise clues that were nearly missed in low-light conditions.

Augmenting DRE with AI

Oculometrix does not replace the officer — it provides an AI co-pilot with objective eyes. Drug Recognition Experts gain neural-network-quantified nystagmus data and ML confidence scores that supplement their clinical assessment.