Senior AI Engineer
Remote · Part-time or Full-time
We're not building an LLM feature - we're building an AI system that executes accounting workflows end-to-end. The hard part isn't generation; it's making the system behave reliably over time. In the financial domain, "almost right" is a liability. You'll be building deterministic wrappers around non-deterministic models.
What we're building
At Tally, we're creating an AI-first accounting platform for SMEs. We automate financial workflows with extreme correctness requirements. This involves:
• Long-running, stateful processes: moving beyond single prompts to multi-week execution loops.
• High-stakes data: structured and unstructured financial records where precision is non-negotiable.
• Systemic reliability: systems that are right by design, not just plausible by chance.
The problems you'll own
• The "demo to prod" gap: agents that handle messy, real-world edge cases.
• State drift: preventing agent degradation across complex, multi-step workflows.
• Predictability: turning non-deterministic LLM outputs into predictable/measurable behaviour.
Examples of things you'll do
Build evaluation frameworks. Design stateful systems. Improve determinism. Instrument for rigor.
What we're looking for
• Systems thinking: experience building stateful AI systems that operate over real, messy workflows.
• Engineering rigor: strong fundamentals in software engineering.
• Iteration mastery: you've lived through agent failure modes and built the loops to fix them.
• Hands-on experience with evals / automated benchmarking, context management and memory architectures, and advanced agent design patterns.
• A metrics-first mindset: if it isn't measured, it isn't solved.
Apply for this role
A short form - we'll reply personally.