Alex Nyberg

Available in Dublin, London and Barcelona

From Business Problemto Working Software.

I help companies ship hard software — from AI-native products to industrial systems that have to work in the real world.

Contract, fractional and advisory engagements

See selected outcomesLinkedIn
SpecificationRequirementsAcceptance testsConstraintsGenerateTestVerifyProductionVerifiedTracedReleasedYou approveYou approve
Intent in. Verified software out. People at the gates.
Trusted on programmes for
  • Ford
  • Cisco
  • Sacramento Regional Transit
  • US Federal Agency
  • Boston-area hospitals
  • Fortune 500 enterprises
Years shipping
15+
Revenue delivered
€M+
Client tier
F500
Zero to live demo
24h

A point of view on applied AI.

  1. Specs before prompts

    Write down what the business needs first. Agents work from a reviewed specification, not a chat history.

  2. People at the gates

    AI does the heavy lifting. People approve the spec, the consequential decisions and every release.

  3. Every result traceable

    Each change traces back to a requirement, so the software stays explainable to stakeholders.

Applied AI engineering that ships business outcomes.

Spec-driven development

From requirements to software, with the spec as the source of truth.

  • A working MVP in days or weeks, not months.
  • Fewer misunderstandings between business and engineering.
  • Software you can explain to stakeholders and auditors.
How it works

Most AI-assisted coding starts from a prompt and ends in code nobody can explain. I work the other way round — capture what the business needs as a clear, reviewable specification, then use LLM agents to generate the architecture, code, tests and infrastructure from it, and check the result back against the spec. It's the approach behind Specatron, the platform I founded, and how I built Market Matcher end to end in a few days.

How AI is used

  1. Turn workshops, documents and conversations into structured, testable requirements.
  2. Generate code, tests and infrastructure-as-code from the spec with multi-agent workflows (Claude Code, sub-agents, MCP).
  3. Trace every change back to a requirement, so reviews and audits stay simple.
  4. Keep people approving the plan, the architecture and each release.
  1. 1CaptureWorkshops and documents become requirements
  2. SpecifyA reviewable, testable specYou approve
  3. 3GenerateAgents write code, tests and infrastructure
  4. 4VerifyEvery change checked against the spec
  5. ReleasePeople approve each releaseYou approve

Business process automation

Taking repetitive, document-heavy work off people's desks, safely.

  • Hours of manual work saved every week.
  • Faster turnaround for customers and colleagues.
  • Fewer manual errors, with a clear audit trail.
How it works

Many processes still run on email, spreadsheets and copy-paste between systems. I design automations where LLMs handle the unstructured parts — reading documents, classifying requests, extracting data, drafting replies — and connect them to the systems you already use, with clear rules for when a person decides.

How AI is used

  1. Map the process and find the steps where an LLM adds value, and the ones where it shouldn't.
  2. Build extraction, classification and drafting steps with structured, validated outputs and retrieval over your own knowledge (RAG).
  3. Integrate with existing systems through APIs, queues and connectors — the same "make systems talk" work I've done across industrial, transit and enterprise platforms.
  4. Human approval on anything consequential, with logging and measured accuracy.
  1. 1CaptureEmails, documents, forms and requests come in
  2. 2UnderstandLLMs read, classify and extract the key data
  3. 3PrepareDrafts replies and actions, checked against your rules
  4. DecideA person approves anything consequentialYou approve
  5. 5ActYour systems are updated, with an audit trail

Projects delivered.

Automotive OEM

Connected-vehicle platform

Took an embedded edge-to-cloud product from zero to MVP to win the contract. It now carries millions of vehicle connections worldwide.

Sacramento Regional Transit

Light-rail monitoring and PA system

In-vehicle agents, on-prem monitoring and system integrations that brought SacRT into regulatory compliance.

US Federal Agency

Infrastructure monitoring

Monitoring software for physical infrastructure, deployed on-premise under strict security requirements.

Market Matcher

Route-to-market tool for B2B SaaS

Built end to end and deployed in a few days using agentic workflows. Live at marketmatcher.co.

Boston-area hospitals

Energy monitoring and control

Transfer-switch monitoring over Modbus and SNMP, unlocking up to $70,000 a day by moving load out of peak-rate windows.

Ford, with Cisco

Natural-language prototype

From zero to live demo in 24 hours. In-vehicle calling and connected traffic lights, demoed at the Detroit Smart Parking Lab.

Products built with speed, accuracy and explainability.

I've spent 15+ years shipping software in complex, fast-moving environments, both hands-on and as a technical leader.

My work sits where software, hardware, people, and business meet. I bring clarity to ambiguity, uncover the business priorities that matter most, and lead products from the first customer conversation through to a production launch.

AI-first engineering.

Turning specifications into production-ready software.

Solve the impossible

The hard, ambiguous, high-stakes problems others walk away from. The one you send in when it has to work.

Bridge business & tech

Fluent in both boardroom and build pipeline, translating commercial goals into engineering decisions.

Operate under pressure

Comfortable delivering business critical work on a deadline. Calm, methodical, and decisive when the stakes are real.

Make systems talk

Connecting platforms, devices, and software of any kind when data needs to flow across boundaries.

Experience.

  1. 2026 — PresentDublin, Ireland

    Applied AI Engineer & Founder

    Specatron

    Started and run Specatron, a platform that takes written requirements and generates working software from them. I own the product and technical direction end to end, from capturing requirements to deploying code.

    Highlights
    • Own the product and technical direction end to end, from capturing requirements to deploying code.
    • Built with Go, PostgreSQL, Directus, and Docker, using Claude's APIs.
    • Handle the business side: defining who we sell to, talking to customers, and closing deals.
  2. 2015 — 2025Dublin, Ireland

    Engineering & Delivery Leadership

    Davra

    Joined as an engineer and worked my way up through three promotions into senior technical and business leadership, covering software development, project management, and delivery for Fortune 500 clients.

    Highlights
    • Hired and led a team of engineers building embedded agents, edge-to-cloud apps, and systems integrations for environments with strict regulatory and security requirements.
    • Responsible for millions in delivered project revenue across automotive, public transit, healthcare, border security, and heavy industry in Europe and the US.
    • Worked alongside sales as the technical lead: ran discovery workshops, built prototypes that won deals, then delivered what we promised.
    • Generated an estimated €1,000+ per hour in value through prototypes and demos that turned directly into signed contracts.
  3. 2013 — 2014Dublin, Ireland

    Technical Founder

    NexEd Labs

    Founded a development and marketing studio serving Irish SMEs as its technical founder. Brought on a co-founder, won clients and built recurring revenue, running both the delivery and the commercial side.

  4. 2009 — 2012Dublin, Ireland

    Technical & Management Roles

    HP Enterprise Services

    Earned three promotions across technical and management roles at HP Enterprise Services, building the operational foundation of my career in process management, service management, and running IT at scale.

Technologies.

The tools and domains behind the outcomes above.

AI engineering

  • Claude Code & sub-agent workflows
  • multi-agent orchestration
  • Anthropic & OpenAI APIs
  • MCP servers
  • LangChain
  • prompt and context design
  • RAG knowledge workflows
  • LLM application architecture
  • neural networks
  • predictive maintenance
  • NLP

Languages & frameworks

  • Python
  • Go
  • JavaScript/TypeScript
  • C/C++
  • Flask
  • Vue.js
  • React

Cloud, platform & DevOps

  • AWS
  • Docker
  • Terraform
  • Kubernetes
  • GitHub Actions
  • CI/CD
  • Cloudflare
  • Linux administration
  • edge-to-cloud deployment
  • on-prem and air-gapped delivery

Embedded, IoT & protocols

  • Yocto Linux
  • Arm64
  • Buildroot
  • industrial gateways
  • MQTT
  • CAN bus
  • J1939
  • Modbus TCP/RTU
  • BACnet
  • SNMP
  • OPC
  • LoRaWAN
  • REST/SOAP

Data & backend

  • PostgreSQL
  • Directus
  • REST APIs
  • GraphQL
  • gRPC
  • Kafka

Observability

  • Grafana
  • Prometheus
  • structured logging
  • alerting and incident response

Leadership

  • Hiring and mentoring engineers
  • technical strategy and architecture
  • delivery and programme management
  • technical pre-sales
  • agile and waterfall
  • GDPR and regulated-sector compliance

Delivery & process

  • Agile
  • Scrum
  • Kanban
  • waterfall
  • regulated-sector delivery
  • security hardening
  • air-gapped deployment

Domains

  • Industrial IoT
  • automotive and connected vehicles
  • transportation and transit
  • manufacturing
  • energy and utilities
  • telecom
  • fleet management
  • critical infrastructure

Available for the right problem

Have a hard thing worth building?

Tell me what you're working on. I usually reply within a day or two.

Email Alex