NAVIGATIONL39 / NAV
// Precision · Discretion · Resilience Close
// Engineering & Applied AI

We build what others only advise.

L39 Tech is a small team of principal-level engineers working across applied AI, industrial hardware and secure software. No account managers, no delivery theatre — the people who scope your project are the people who build it.

Free 20-minute technical scoping call — talk to an engineer, not a salesperson. Book a slot →
// Capabilities

Five disciplines. One team.

We are deliberately broad, because most real problems do not respect the boundary between software, hardware and data. Every engagement is delivered by senior engineers with end-to-end ownership.

Applied AI Systems

Models that survive contact with production. We build retrieval systems, computer vision pipelines and decision-support tooling — then own the evaluation harness, the drift monitoring and the cost envelope that keep them useful.

LLM Systems Computer Vision Edge Inference MLOps

Industrial IoT & Hardware

Sensors, gateways and firmware for environments that punish shortcuts. Schematic capture through to PCB fabrication, enclosure, certification support and fleet telemetry — designed for a ten-year service life, not a demo.

Embedded C RTOS LoRaWAN PCB Design

// In-house PCB manufacture — same-day prototypes, no external fab lead time.

Secure Software & Apps

Platforms built for regulated and adversarial environments. Threat-modelled architecture, hardened build chains, auditable data handling and the documentation your assessors will actually ask for.

Zero Trust Cloud Native Cyber Essentials+ Secure SDLC

SME & Startup Strategy

Fractional CTO work and technical due diligence for teams that need judgement rather than headcount. Build-versus-buy, architecture review, hiring plans, and honest answers about what your platform can carry.

Fractional CTO Technical DD Architecture Review

End-to-End Prototyping

Concept to working hardware-and-software prototype, typically in weeks. We prove the difficult part first — the sensor, the model, the latency budget — so the commercial decision is made on evidence.

Rapid Prototyping Proof of Concept Feasibility
// Selected Work

Real problems. Real outcomes.

A selection of projects that demonstrate what precision engineering looks like in practice. Some clients are named; others are not, and never will be.

// Anonymised Client
Defence Firmware & Secure Software

Counter-UAS Firmware, Electronics & Software

Three years as engineering partner to a UK counter-drone technology company, building custom firmware, control electronics and software for systems used by military and law enforcement customers.

  • Began as one component, grew to the full technical stack
  • Now one of the client's primary engineering contractors
  • Security-first development from the ground up
Consumer Products Prototyping

AME Group — Rapid Prototyping Partner

Proof-of-concept and prototyping support across a diverse consumer product range, from a misting fruit bowl to a sound-embedded yoga mat — embedded electronics, sensors, custom PCB and software, whatever the next idea needs.

  • One product now launching
  • A second in overseas manufacturing, pre-launch
  • A third in active development
// The People

Senior engineers, not agencies.

Every engagement is staffed by people who have shipped the thing you are asking for — in production, under constraint, with their name on it.

// Patent Pending

Project Cerberus

Cerberus is L39's patent-pending edge AI system for acoustic drone detection. It classifies a live acoustic data stream on the device itself, using multiple AI models rather than depending on a single classifier.

Running inference at the edge means the decision is reached where the microphone is, without a round trip to a datacentre and without depending on a network link being there at all.

Development is ongoing, and early testing has put numbers against it.

TRL 5 Technology validated in a relevant environment
300–400 m DJI Phantom 4 · line of sight, open field · early testing
Acoustic / Live Stream Continuous audio captured and processed in place, rather than recorded and shipped elsewhere for analysis.
Models / Classification The stream passes through several AI models in sequence, each contributing to the classification rather than one deciding alone.
// Contact

Tell us what you are building.

Send the problem, not a brief. If it is outside what we do well, we will say so and point you at someone better. No pitch decks required.

Prefer to talk first?
20 minutes with an engineer. No obligation, no sales script.

Book a Call
// Typical response: same working day