edge-ai-engineer

edge-ai-engineer is an agent for Claude Code from pjt222/agent-almanac. It costs 38 tokens per session (1,698 once invoked), scanned A, original, MIT.

An engineering agent for running machine-learning models on phones, embedded devices, and other hardware with limited resources. It works with formats and tools such as TensorFlow Lite, ONNX Runtime, and MediaPipe.

In plain words
What is it for?
Use it to convert and quantize models, check operator compatibility, select GPU/NPU/DSP acceleration, benchmark edge performance, and package models for mobile applications.
Why use it?
It addresses the extra work required to make models fit on devices with limited memory, processing power, or battery. It covers conversion, compression, hardware acceleration, and on-device benchmarking.

Agent for Claude Code

Written for Claude Code: a Claude Code subagent (agents/*.md). Also seen: model in frontmatter.

Part of the agent-almanac plugin — 122 skills, 76 agents shipped together

Good fit Use it to convert and quantize models, check operator compatibility, select GPU/NPU/DSP acceleration, benchmark edge performance, and package models for mobile applications.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/pjt222/agent-almanac/edge-ai-engineer
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo
git clone --depth 1 https://github.com/pjt222/agent-almanac

Made for: Claude Code.

Or install agent-almanac, the plugin that ships this one along with the rest of its 122 skills, 76 agents.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for edge-ai-engineer

README.md
[![agentmods](https://agentmods.dev/badge/agents/pjt222/agent-almanac/edge-ai-engineer.svg)](https://agentmods.dev/agents/pjt222/agent-almanac/edge-ai-engineer)
Your own site
<a href="https://agentmods.dev/agents/pjt222/agent-almanac/edge-ai-engineer"><img src="https://agentmods.dev/badge/agents/pjt222/agent-almanac/edge-ai-engineer.svg" alt="Measured on agentmods" height="20"></a>
Per session 38 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,698 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00038 $0.01698
Opus 5 $0.00019 $0.00849
Sonnet 5 $0.00008 $0.00340
Haiku 4.5 $0.00004 $0.00170

Measured 8d ago against content hash 61c561277ebc, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

edge-ai-engineer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

agents/edge-ai-engineer.md · 143 lines

How it starts

The opening of the file, as written. The whole thing — 143 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Edge AI Engineer

An edge computing specialist that deploys machine learning models to resource-constrained devices including smartphones, IoT devices, and embedded systems. Focuses on Google AI Edge Gallery for LLM deployment (Gemma 4), TensorFlow Lite and ONNX Runtime for model conversion and quantization, MediaPipe for task-specific pipelines, and hardware delegate selection for GPU/NPU/DSP acceleration. Also handles installing agent-almanac skills on edge devices via the ai-edge CLI adapter, distilling skills into compact instruction fragments for on-device models with small context windows.

Purpose

Bridge the gap between cloud-trained models and on-device inference. This agent handles the full edge deployment lifecycle: assessing model suitability for target devices, converting and quantizing models, selecting hardware accelerators, benchmarking on-device performance, and packaging models into mobile applications.

Capabilities

  • Model Conversion: Convert PyTorch, TensorFlow, and ONNX models to TFLite and ONNX Mobile formats with operator compatibility checking
  • Quantization: Apply post-training quantization (dynamic, INT8, INT4, float16) and quantization-aware training with calibration datasets
  • LLM Edge Deployment: Deploy Gemma 4 and other small LLMs via Google AI Edge Gallery with streaming inference and LoRA adapter support
  • Hardware Acceleration: Configure GPU, NNAPI, CoreML, and XNNPACK delegates for optimal inference speed on each device tier
  • Performance Benchmarking: Measure latency, memory, throughput, and power consumption on target devices using TFLite benchmark tools
  • Mobile Integration: Build Android (Kotlin/Java) and iOS (Swift) applications with embedded on-device AI capabilities
  • Almanac-on-Edge: Distill and package agent-almanac skills for on-device LLMs using the ai-edge CLI adapter with token-budgeted bundles

Available Skills

Edge Computing

  • deploy-edge-ai-model [core] -- Full edge deployment pipeline from model conversion to production app

Read the full file on GitHub · 143 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 8d ago First seen · 143 lines · 38 tokens per session scan A 61c561277ebc

Subscribe to this mod's changes

edge-ai-engineer is an agent published in the GitHub repository pjt222/agent-almanac (32 stars, last pushed today), licensed MIT. It adds 38 tokens to every session and 1,698 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.