Engineering services
AI-augmented engineering team. Federal-grade work. Startup speed.
Rapid prototyping, system engineering, and demo development.
info@akmee.net · akmee.net
What we do
Pick one, pick all three. Each engagement is led by the principal and executed by the fleet.
01
From a spec, a sketch, or a napkin idea to a working demo in days — not weeks. We run research, design, build, and test in parallel so you see something real fast.
02
Full-stack from bare metal to cloud. Hardware bring-up, firmware, backend services, data pipelines, and front-end — designed as a system, not bolted together.
03
Investor decks are fine. A live demo that does the thing is better. We build demos that hold up under scrutiny — real data, real hardware, real performance.
Engagement models
Pick the model that fits your team, timeline, and how much you want to own the process.
AK-mee runs the engagement end-to-end.
Dedicated fleet operating under your direction.
Stand up your own AI engineering fleet.
How we work
Each project is led by Chad Smith (the principal) — the engineer who sets architecture, owns the client relationship, and signs off on every deliverable.
The fleet — a roster of specialist AI agents — handles implementation slices in parallel: research, build, test, security review, communication. Agents don't replace engineering judgment; they multiply it. The principal decides what gets built and how it gets tested. The fleet ships it.
The result is small-team accountability at large-team velocity. You get one point of contact, not a committee. And deliverables that move at the speed of a well-run sprint, not a quarterly roadmap.
The fleet runs on the same infrastructure we build for clients — dog-fooded, battle-tested, and running live in the lab right now.
Case studies
Seven projects across embedded hardware, AI inference, federal data pipelines, and custom tooling.
01
Superbird resurrection — custom Matrix UI on a discontinued Spotify display.
Spotify killed the Car Thing (Superbird) in 2024. We brought it back with a custom Matrix-themed web app, Spotify playback control, fleet messaging overlay, and a full device simulator. The 480×800 capacitive screen now runs a living UI instead of a brick. ADB sideload, over-the-air deploy pipeline, and a BT PAN bridge to the home lab network.
02
Small-form display module showcasing THE RAT brand identity v8.
A compact LCD display project built to push the v8 brand identity refresh in physical form. Custom enclosure (FDM-printed, ABS), embedded display driver, and the full matrix rain + avatar composite running on constrained hardware. Served as the reference for the display subsystem now deployed across the fleet.
03
High-speed recording benchmark with fleet-wide results dashboard.
Benchmark harness for high-speed video recording across heterogeneous hardware. Multi-block-size sweep, fan-controller integration (17V DC bus, custom firmware), real-time telemetry, and a web dashboard that aggregates results across the fleet. Surfaced a 23% throughput gap between storage configs that changed the hardware spec.
04
Phone companion for monitoring and driving a multi-machine AI fleet.
Real-time session monitor for a fleet of Claude Code agents running across multiple machines. Tile grid, terminal pane with live hook events, mic + waveform + transcription, kiosk overlay, HTTPS via Tailscale. Built so the principal can manage 6+ concurrent AI engineering sessions from a phone while walking the lab.
05
Matrix-themed AI avatar — voice interface for the workstation fleet.
OpenAI TTS (onyx, 1.2x), Matrix rain, CRT flicker, Spotify control, fleet-wide message overlay, MQTT pub/sub, SSE event bus, and a headless CadQuery / OpenSCAD pipeline. The front-end for every AI interaction in the lab — runs on a Car Thing, browser, and terminal simultaneously. Project-managed and built by the same agent fleet it manages.
06
High-speed video recorder with AI-driven pan-tilt-zoom on Jetson Orin.
YOLO11 + TensorRT object detection pipeline driving a PTZ camera on an NVIDIA Jetson AGX Orin (128 GB RAM). Latency-optimized inference, sub-100ms track-to-slew, configurable target-class filtering, and a web control panel. Demonstrated live at a trade show — camera acquired and tracked targets across a 15-meter floor autonomously.
07
Fleet-as-WiFi-locator — RF multi-tool integration with distributed RSSI sensing.
Flipper Zero (RF multi-tool) attached to the fleet hub (box), combined with distributed WiFi scanning across four fleet members (box, dart, jetson, octo). Each node runs iw dev scan via SSH on demand; a central server aggregates per-host SSID/BSSID/RSSI readings and exposes them via a REST + SSE API. Multi-vantage RSSI gradients constrain transmitter position. New-device alerts broadcast to all Dude fleet instances in real time. Active R&D — demonstrating fleet-as-positioning-network pattern.
08
AR sandbox with real-time gesture control and terrain visualization.
Augmented-reality sandbox: Arducam ToF depth sensor + pico projector on a Jetson platform, real-time contour rendering from live point-cloud data, gesture recognition for terrain interaction. Custom FDM enclosure (ABS), Makefile-driven bring-up, containerized vision pipeline. Physical terrain that talks back.
Team
One principal. Ten specialist roles. Every engagement gets the full roster.
Principal
Leads every engagement. Owns architecture, client relationship, technical direction, and final sign-off. The principal is accountable for every deliverable — not a coordinator, but the engineering lead.
Project management
Plans the build, files issues, dispatches the implementation fleet, verifies outputs before they ship. Owns the task board and keeps parallel workstreams from stepping on each other.
Build engineering
Writes the code, assembles the hardware, runs the build. Specialists spin up per codebase — one per codebase at a time to avoid merge conflicts and context collapse.
QA & testing
Regression test runner. Area-gated test suites with mandatory green-gate before any ship signal. No principal-clicks-to-verify; the QA agent asserts the behavior.
Research
Deep-research tasks: prior art, competitive landscape, technical feasibility, regulatory requirements. Drops structured reports the build team works from.
Security review
SAST scanning, code review for vulnerabilities, dependency audit. Runs on every ship candidate. Findings go to the principal before the client sees the build.
DevOps & infrastructure
Container builds, deployment pipelines, service watchdogs, fleet health monitoring. Keeps the CI/CD loop tight and alerts the principal when something drifts.
Compliance & spec enforcement
FAR / DFARS clause review, audit-trail enforcement, formal process adherence. Every deliverable is checked against the spec — not approximately, exactly.
Operations & scheduling
Coordination, scheduling, logistics, and handoffs across workstreams. Keeps the engagement moving without requiring the principal to track every moving part.
Legal counsel
Entity formation, federal contracting clause review, IP / trademark, NDA review, data privacy, and terms-of-service. Research and process scaffolding; not a licensed attorney — consequential decisions get reviewed by outside counsel.
Business development
Identifying opportunities, structuring engagements, and building client relationships. Understands the transactional landscape and moves in it with confidence.
Contact
Tell us what you're working on. We'll tell you what the fleet can do with it.