AI Engineer
Build the AI that answers questions about Saudi companies, where every number has to be right and every answer traces back to its source.
About Q4
Q4 is building research infrastructure for the Saudi market: earnings calls, transcripts, filings and financials for Tadawul-listed companies, in one workspace where every answer traces back to its source. We are a small team in Riyadh, and the people who join now shape how the product works.
Where you’ll work
Fully remote, from Macedonia. You will work with the team in Riyadh, with regular overlap during the working day.
What you’ll do
Own the AI chat end to end: retrieval, prompting, tool use, and how answers are grounded and cited.
Improve retrieval over filings, financial statements and call transcripts, in both Arabic and English.
Build the extraction that turns PDFs and scanned reports into structured, verified financial data.
Set up evaluations so we know when an answer got better or worse, and catch wrong numbers before users do.
Keep model cost and latency under control as usage grows, and choose the right model for each job.
Work on Q4 MCP, the server that brings Q4 into Claude, ChatGPT and Cursor.
Ship to production often, and talk directly with the analysts who use what you build.
What we’re looking for
You have shipped LLM features to real users, not only prototypes.
Strong in TypeScript or JavaScript (Node.js). Python is a plus.
Hands-on experience with retrieval: embeddings, chunking, hybrid search, reranking.
You test AI systems seriously and don’t stop at “looks good in the demo.”
You care about precision. This is financial data, and approximately right isn’t good enough.
Clear written English, and comfort working remotely with a lot of ownership.
Based in Macedonia.
Nice to have
Experience with Arabic NLP or other multilingual systems.
Document parsing and OCR, especially tables in financial reports.
Familiarity with financial statements or capital markets.
Experience with MCP, agents or tool-calling frameworks.
PostgreSQL and Prisma.
How we work
Small team, short feedback loops, and very little process between an idea and something users can try.
You will work directly with the founders rather than through layers.
We care about precision. This is financial data, and being approximately right is not good enough.
Apply
Applications for this role go through a short form: your CV, LinkedIn, where you’re based and something you’ve built. It takes about five minutes.
Apply for this role