# SIGINTENT — Antonio Elena

> Independent technology advisory: fractional CTO, architecture review, AI strategy, and cloud and distributed-systems guidance. Taken on selectively alongside a CTO role at BEAI Energy. The central stance is that most technology decisions are framed wrong before they are answered, and that fixing the question is where senior judgment earns its fee.

Antonio Elena has spent two decades across enterprise, consulting and startups as an architect, engineering leader and CTO. He was previously Global Head of Architecture & Technology at SGS, reporting to the Group CIO and running the architecture function across 800+ projects. Past work includes KPMG, Aviva, Maersk, Experian, Equifax, Siemens and PwC.

Engagements: fractional and interim CTO; architecture review; AI strategy sprint; advisory retainer. The differentiator is execution capability — the strategy can be carried into delivery rather than handed over as a deck.

## Essays

- [The evolution of the software architect: from builder to system designer](https://sig-intent.com/writing/evolution-of-the-software-architect/): The architect's job moved from drawing diagrams to framing decisions. What the elevator model leaves out is what the ride costs to keep.
- [M-shaped engineering: cheap breadth is why you need more than one depth](https://sig-intent.com/writing/m-shaped-engineering/): Why pure specialization fails when AI makes breadth cheap — and why the naive generalist argument fails too. The second depth is what changes anything.
- [MD, JSON, HTML or XML: pick the format for the next reader](https://sig-intent.com/writing/pick-the-format-for-the-next-reader/): Most format arguments are really arguments about who reads the artifact next. XML for contracts, HTML for the interface, Markdown for the conversation.
- [AI is an architecture problem, not just a model problem](https://sig-intent.com/writing/ai-is-an-architecture-problem/): Shipping AI in production is a distributed-systems problem with ML as one component. Why POCs stall at the demo-to-production cliff, and the reframe.

## Site

- [Writing archive](https://sig-intent.com/writing/): All essays, newest first.
- [Full text of every essay](https://sig-intent.com/llms-full.txt): The whole corpus in one file.
- [RSS feed](https://sig-intent.com/feed.xml): For subscribing.
- [Engagements](https://sig-intent.com/#engagements): The four ways of working together.
- [Approach](https://sig-intent.com/#approach): Diagnose the real problem, reframe the decision, commit to a recommendation.

## Elsewhere

- [aelena.com](https://aelena.com): Personal hub.
- [Self & System](https://selfandsystem.substack.com): A separate publication on psychology, philosophy and AI. Different register, same author.
- [trustworthy-ai](https://github.com/aelena/trustworthy-ai): A working body of knowledge for in-house AI assurance — evaluation, bias, differential privacy, adversarial testing, agentic safety and the regulatory frame.
- [Build a Tokenizer in C](https://aelena74.gumroad.com/l/c-tokenizer): Book. Builds OpenAI's tiktoken tokenizer from scratch in C, base64 decoding up to a complete encoder, over eight chapters. The findings are part of the text: a regex pattern that was wrong on half the inputs, a use-after-free in the arena allocator, and a performance number nobody had measured.
- [XML-Structured Prompting](https://aelena74.gumroad.com/l/xsp): Book. Prompting treated as a software engineering discipline.
- [github.com/aelena](https://github.com/aelena): Code and tools.