What we test for
Typical work
- Retrieval over your own documents, with answers that cite their sources
- Hybrid search and reranking, for when plain vector search stops being enough
- Structured extraction from contracts, tickets, invoices and scanned PDFs
- Taking a prototype from "it demos well" to an accuracy you would publish
The coding test
Index a messy document set, scanned pages and tables included, and answer questions over it with citations. Then we change the chunking strategy on you and ask what the number did and why. Scored on retrieval quality before anything else, because the model is rarely the part that is wrong.
Five interview questions you can reuse
We use questions like these in our soft-skills and technical rounds. Steal them for your own interviews.
Vector search returns the wrong passage. What are your options, in order?
A strong answer covers: Look at the chunk first: size, overlap, whether the heading survived. Then query rewriting, then hybrid search with BM25, then a reranker. Changing the embedding model is a bigger move than people think, because it means reindexing everything.
How do you chunk a document that has tables and headings?
A strong answer covers: Along the structure rather than every N characters: a table stays whole, a heading travels with its section, and every chunk carries enough metadata to be cited. Fixed-size splitting is the default that quietly destroys the hardest documents.
The model answers confidently from nothing. What do you do?
A strong answer covers: Make "I do not know" a scored outcome rather than a failure. Require citations and check them, threshold on the retrieval score, and show the sources so the reader can judge. A fluent wrong answer is worse than no answer.
When do you fine-tune instead of retrieving?
A strong answer covers: Retrieval for knowledge that changes, fine-tuning for a format or a behaviour the prompt cannot hold. Most teams asking to fine-tune have a retrieval problem, and fine-tuning freezes the knowledge at the moment you trained.
How do you keep latency down with retrieval in the path?
A strong answer covers: Measure the stages separately: embedding, search, rerank, generation. Then cache the stable prefix, retrieve fewer and better passages, stream the first token early, and put a timeout and a fallback on every network call.
One monthly price per developer. Everything in it.
No deposit, no minimum term, no placement fee. You pay one invoice a month for each developer on your team.
Per developer, all-in. Your quote arrives by email with your exact number after you send the brief.
- Developer pay, local payroll and taxes
- Vetting: coding test, soft skills, English
- Free replacement within 90 days
- A direct line to our CEO
What changes the price
- Seniority and role. Senior engineer, tech lead or architect.
- Stack. Some stacks and niche tools cost more to staff.
- Team size. Larger teams get a better monthly rate per seat.
- White-label. Agencies reselling under their brand get partner pricing.
Terms in plain words
- First 2 weeks risk-free: if it isn't working, you don't pay for them
- Month to month, 30 days' notice to end
- No deposit and no minimum number of seats
- Want to hire the developer full-time? Buyout of 3 months' fee in year one, 2 months' after
Questions tech leads ask us
How senior are your LLM developers?
Six or more years as software engineers, and at least one retrieval system that reached production and survived real documents. Nobody has six years of LLM work, and anyone claiming it is telling you something useful.
Can I see the coding test results?
Yes. Every profile card carries the score and the reviewer notes, including how their retrieval scored on our question set.
What does it cost?
$6,900 per developer per month, all-in, and flat: the same rate whatever the stack. Your exact quote arrives by email after you send the brief.
Skip the sales rep. Talk to the CEO.
Tell us what you're building, your stack and when you need people. Three matched profiles within 48 hours, and a straight answer on whether we're the right fit.