// Decision Engine Platform

Design, plan, and estimate edge AI deployments

Three primary engines for the core deployment tasks. Supporting engines for subsystem-level validation. All deterministic, all with JSON export.

Primary Engines

Start here. Each engine addresses a core deployment task and orchestrates the supporting engines it needs.

System Designer

Design a complete edge AI system architecture. Describe your deployment once — get a platform recommendation, architecture diagram, and compute sizing with risk flags and tradeoffs.

Deployment profile Camera count AI task Power / cost constraints Optimization goal

→ Architecture recommendation · Risk flags · Tradeoff comparison · Alternatives

Powers: Inference Estimator · Memory Estimator · Stream Calculator · Model Comparator

Full Deployment Planner

Plan infrastructure requirements and deployment feasibility. Integrates hardware selection, power infrastructure, network sizing, storage endurance, and cost estimation into one validated, exportable BOM.

Scenario Camera count Resolution Model Power infra Retention

→ Complete BOM · Cost breakdown · Design flags · JSON spec

Powers: Hardware Selector · GPU Sizing · Module Power · Power Budget · Network Bandwidth · Storage Endurance

Estimate hardware and operational deployment costs. Itemized BOM, per-camera economics, power consumption, and annual energy for any edge AI configuration.

Accelerator family Camera count Configuration Storage type Power infra

→ BOM · Cost per camera · Annual energy · Confidence · Export

Supporting Engines — Deployment Planning

Subsystem-level engines for infrastructure sizing. Use independently or let the primary engines orchestrate them.

Hardware Selector

Choose the right accelerator family — Jetson, Coral, Hailo, or RK3588 — based on scenario, model type, stream count, power budget, and environment.

Scenario Model type Stream count Power budget Environment

→ Ranked platforms · Confidence score · Purchase links · JSON export

Validate compute headroom for inference workloads. Calculates required TOPS, projected FPS, and latency against platform capabilities.

Model architecture Batch size Latency target Throughput

→ Platform match · FPS estimate · Headroom risk · JSON

Module Power Calculator

Size node power budgets and infrastructure. Calculates TDP, PSU specification, cable gauge, battery capacity, and runtime for Jetson, Coral, and Hailo modules.

Module family Operating mode Workload profile Battery runtime Peripherals

→ Module TDP · PSU spec · Battery capacity · Current draw · JSON

Power Budget Planner

Size total system power draw, PoE switch budgets, battery capacity, or UPS runtime. Flags devices exceeding power class limits.

Device count PoE class UPS runtime Cooling load

→ Total TDP · PSU spec · Switch recommendation · UPS sizing · JSON

Network Bandwidth

Model bitrate for multi-camera deployments. Sizes network capacity and flags saturation risk.

Camera count Resolution Encoding Uplink budget

→ Total Mbps · Switch spec · Uplink recommendation · JSON

Storage Endurance

Model NVMe wear for continuous recording. Calculates drive lifespan based on workload, write-amplification, and retention needs.

Bitrate Retention period Drive TBW Write amplification

→ Capacity needed · Drive spec · Lifespan · JSON

Supporting Engines — Model & Inference

Specialized engines for validating model performance, memory fit, and throughput on target hardware.

Inference Estimator

Lookup model performance on target hardware. Queries vendor benchmarks, interpolates missing data, and applies pipeline overhead.

Hardware platform Model variant Precision Runtime engine Resolution

→ FPS · Latency · Confidence · Benchmark source · JSON

Memory Estimator

Validate VRAM and system memory fit. Accounts for weights, activations, runtime overhead, workspace, and OS headroom.

Hardware Model variant Precision Runtime engine Batch size

→ Total memory needed · Memory tier · Unified fit · Warnings · JSON

Stream Calculator

Determine maximum simultaneous camera streams on your target hardware at desired FPS, accounting for pipeline overhead and memory impact.

Hardware Model & precision Target FPS Resolution Pipeline overhead

→ Max streams · Feasibility flag · GPU utilization · Headroom · JSON

Model Comparator

Compare model architectures across accuracy, latency, memory, and throughput. Helps choose the best fit for your constraints.

Model families Hardware target Precision Batch size Priorities

→ Ranked models · Accuracy vs FPS · Memory vs latency · Recommendations

// How to choose

Finding the right engine for your task

Start with a primary engine for end-to-end results, or use a supporting engine for subsystem-level validation.

Engine	Primary Goal	When to Use
Primary Engines
System Designer	Architecture recommendation with risk flags	Designing a new edge AI system from scratch
Deployment Planner	Complete specification and BOM	End-to-end project planning and validation
Cost Estimator	Full BOM and per-camera cost	Budgeting and RFQ preparation
Supporting — Deployment Planning
Hardware Selector	Platform recommendation	Comparing Jetson, Coral, Hailo, RK3588
GPU Sizing	Compute headroom and FPS estimate	Validating inference performance on target hardware
Module Power Calculator	Per-node power budgets	Sizing PSU and battery for individual nodes
Power Budget Planner	PSU, PoE, UPS specification	Sizing power infrastructure for multi-node deployments
Network Bandwidth	Multi-camera bitrate and switch spec	Sizing network capacity for edge stations
Storage Endurance	NVMe drive selection and lifespan	Planning 24/7 recording storage
Supporting — Model & Inference
Inference Estimator	Model FPS and latency	Looking up performance benchmarks
Memory Estimator	VRAM and system memory required	Validating memory fit on target hardware
Stream Calculator	Node count for multi-camera workloads	Sizing deployment for specific FPS and camera count
Model Comparator	Ranked model selection	Choosing best model architecture for your constraints