Software engineering for fintech — and the AI systems being built on top of it.
EOF Systems works with financial platforms on the parts that have to be right: how money moves, how state is reconciled, and increasingly, how AI-assisted workflows can be trusted inside a regulated product.
Where I’m usually brought in.
Most engagements are some combination of two or three of these, over months rather than weeks.
Financial platforms
Payment flows, ledgers, reconciliation and the transactional core of financial products — systems where being approximately right is the same as being wrong.
AI-assisted workflows
Multi-step and multi-agent systems built into real products: orchestration, tool use, evaluation and the guardrails that make model output safe to act on.
Applied AI in regulated products
Bringing LLM-backed features into environments that demand auditability — deterministic boundaries around non-deterministic components, and a paper trail that survives review.
Integration engineering
APIs, webhooks and third-party services built with the idempotency, retry semantics and failure handling that production actually demands.
Modernisation
Inherited systems brought back under control, or platforms migrated between stacks — incrementally, without betting the company on a rewrite.
Embedded engineering
Longer engagements inside a client team, typically as technical lead on a defined area of the platform rather than a detached project.
How I work.
Domain before code
Financial products aren’t a vertical you pick up mid-project. Knowing why a reconciliation runs the way it does, or which failure a regulator will actually ask about, changes what you build long before it changes how you build it.
Non-determinism is a design constraint
Model output belongs behind the same boundaries as any other untrusted input. The interesting engineering in an AI feature isn’t the prompt — it’s the evaluation, the fallbacks, and knowing which decisions a model should never be allowed to make alone.
Incremental over ambitious
Most systems don’t need rewriting. They need a sequence of changes that each ship safely and leave things better. Big-bang migrations usually defer risk rather than remove it.
Embedded, not external
Same repos, same reviews, same standups. Consultants who work at arm’s length produce documents. I’d rather produce running code.
The practice.
EOF Systems is an independent software engineering practice working with fintech teams.
The work has always been transactional systems — the places where correctness and uptime are the requirement rather than an aspiration. More recently that’s extended into applied AI: multi-agent workflows, tool orchestration, and the question of how you put a non-deterministic component inside a product that has to be auditable.
Based in Bangkok, working remotely with teams across several time zones.
Working on something that has to be right?
Tell me what you’re building and where it hurts. I’ll tell you honestly whether I’m the right person for it.
hello@eofsystems.com