Founding AI Engineer
Locations
Top 0.1% engineer. FAANG pedigree or equivalent proof. Operator of software factories, not a passenger in them. Hybrid in Warsaw (WeWork Hotel Europejski, 2 days/week). Top-of-market cash + substantial equity.
About Gralio
Gralio is the context layer for enterprise AI agents.
We capture how knowledge workers actually do their jobs — via lightweight desktop recording — and turn that raw signal into process maps, SOPs, and ready-to-automate workflows using frontier vision models. The thesis is simple: agents are only as good as the context you feed them, and nobody has clean, structured context on real enterprise work. We do.
We are past the "interesting demo" phase:
We are embedded in a major pilot with Deel, a $12B+ HR/payroll platform. The engagement has produced hundred of SOPs, tens of thousands of hours of captured process, and positions Gralio as the supervisor-layer infrastructure behind their agent rollout.
We are running parallel deployments in European BPO and professional-services accounts where we combine process capture with actual automation delivery.
We are venture-backed, cash-efficient, and run by two founders who've been through it before.
The headline is that every serious enterprise rolling out agents over the next 24 months will need a system of record for how work actually happens — and we intend to be that system.
Founders:
Michał Kaczor (CEO) - ex-Bain, product and ops operator across London and NYC.
Tymon Terlikiewicz (CTO) - previously CTO at Batmaid (Swiss)
Why this role exists
The bottleneck at Gralio is no longer "can we build it." It's how fast one extraordinary engineer, leveraged by coding agents, can ship production systems that used to require a team of ten.
We want someone who already operates that way — who runs a personal software factory, delegates aggressively to Claude Code / Codex / equivalents, reviews and merges at a velocity that looks unreasonable to people who still hand-type every line, and has the taste to know when to slow down and get the architecture right.
This is an early engineering hire outside the founding team. Scope is unbounded: the Rust desktop recorder, the TypeScript/SvelteKit backend, the vision-LLM processing pipeline, the customer-facing automation deliverables - all on the table.
Who we're looking for
If you are a strong engineer but not in the top fraction of strong engineers, this is not the role — and we will tell you that honestly.
The baseline:
FAANG, top-tier scale-up, or elite startup background — or an equivalent proof of having shipped at the level those environments select for. We care about the evidence of excellence, not the logo itself. A killer open-source record, a hard product you built solo, or a demonstrable track of outperforming a strong peer group all count.
You've shipped non-trivial software to real users and can walk us through decisions, trade-offs, and failures in specifics — not abstractions.
You operate a software factory. You don't write code the way engineers wrote code in 2022. You run multiple agents in parallel, you've built your own tooling around them, you know where they fail and how to compensate. If your throughput hasn't meaningfully multiplied in the last 18 months, you're not the person we're looking for.
High agency. You see a gap, you close it. You do not wait for tickets. You do not ask permission to fix something obviously broken.
Taste. You can tell the difference between a system that works and a system that is good — and you care about the second one.
Calm under ambiguity. Early-stage means specs change, customers surprise you, and the right answer on Monday is wrong by Thursday. This energises you.
Skill requirements
Must have - non-negotiable:
TypeScript at depth. You've built and maintained real systems in it; you know the ecosystem, the pitfalls, the performance profile, and the patterns.
Agentic coding as a daily practice. Claude Code, Codex, Cursor agents, Aider, or equivalent — you don't just "use AI," you architect your workflow around it. You have opinions about sub-agents, parallel worktrees, and how to structure a codebase so agents can navigate it.
LLM systems in production. You've shipped things that call models in the critical path; you've dealt with non-determinism, built evals, managed cost and latency.
Frontend competence. React, Svelte/SvelteKit, Vue, or similar. Frontend is not something you outsource.
Production discipline. E2E testing, observability, CI/CD — you ship with the lights on.
Strong plus:
Rust at working proficiency or better — enough to own meaningful features in our Tauri-based desktop recorder, debug native-layer issues and review Rust PRs with confidence.
Vision LLMs (Gemini, Claude, GPT-5-class) in applied pipelines.
Cloud infra at the level of owning it (GCP).
Prior founding-engineer or early-startup experience at a company that went somewhere.
Current stack
Subject to change as we scale — but today:
Desktop: Rust + Tauri, macOS + Windows, custom capture layer.
Backend & frontend: SvelteKit, TypeScript end-to-end.
Infra: GCP, Cloud Run, Postgres.
Location, comp, contract
Location: Hybrid — Poland-based, 2 days/week in-person at WeWork Hotel Europejski, Warsaw. The rest is yours to arrange.
Contract: B2B, USD.
Comp: Top-of-market cash base plus substantial founding-engineer equity. We do not compete on being slightly cheaper than Big Tech - we compete on outcome and ownership.
Gralio values
Meritocracy. The best idea wins. Seniority is not a tiebreaker.
User obsession. We sit next to real operators and watch them work. The answer is always in the workflow.
Empowerment through information. Everyone sees the full picture - revenue, pipeline, product decisions. We hire adults.
Velocity with judgment. Move fast on what's reversible. Slow down and get it right on what isn't.
Explorers. We'd rather try something and fail than do something boring and succeed.
Process
Not set in stone, but the shape is:
15-min screen (apologies if this is AI based, we're AI first after all)
Deep technical conversation - your past work, hard problems, how you think. No whiteboard trivia.
Paid work trial - typically one full day, often more, building something real with the founding team. This is how we both decide.
References - we will call people you've shipped with. We expect the same of us.
We move fast. If you're the right person, you'll have a decision in days.
Apply
Send your CV and - more importantly - one piece of evidence you're exceptional. A repo. A thing you built alone. A story of a problem you solved that others couldn't. A prompt-engineering system you're proud of. The weirdest, most specific artifact you have.
