IREVEX

AI products you can check before you trust

IREVEX is an AI startup. Our first product, Beyond Queries, answers questions about your data in plain English and shows its working before anything runs. We also build AI systems for other teams.

Try Beyond QueriesWork with us

Which customers churned after a failed payment?

select c.id, c.email, p.failed_at
from customers c
join payments p on p.customer_id = c.id
where p.status = 'failed'
where p.status = 'failed' and c.cancelled_at > p.failed_at

Churn means cancelled after the failure, not merely a failed payment.

Synthetic schema and figures.

About IREVEX

We build our own AI products, and we take on a small number of engineering projects for other teams. Same engineers, same standards. Ordinary software fails deterministically. AI fails statistically. It is right most of the time, and the failures look exactly like the successes. Building for that is the whole job.

Our product

Beyond Queries

Ask your warehouse a question in plain English. Beyond Queries writes the SQL and puts it in front of you before anything executes, so a wrong definition is visible rather than buried in a chart.
01

Ask in plain English

“Which customers churned after a failed payment?” No SQL, no dashboard to build first.

02

See the query it wrote

It reads your real schema and your settled definitions, then shows the SQL in full.

03

Approve, then it runs

Nothing touches your warehouse until a person has read the query and said yes.

Open Beyond QueriesHow it worksLive at beyondqueries.com · onboarding a small number of teams
  • Python
  • FastAPI
  • LangGraph
  • PostgreSQL
  • Qdrant
  • vLLM
  • Redis
  • AWS
  • Docker
  • Celery
Services

We also build for your product

The engineering behind Beyond Queries is available for your system too: agents, retrieval and the infrastructure that keeps them running. You talk to the engineer who would do the work, on the first call.

Book a callHow engagements work

Where AI systems break

We work here

Interface
Chat · API · scheduled jobThe request that never comes back
Harness
Routing · retries · limits · permissionsThe retry storm, the runaway spend
Tools
Typed · permissioned · tracedThe action nobody authorised
Systems
Your database · your APIsThe write that cannot be undone
01

AI agent development

Multi-step agents with tool use, delegation and a trace for every decision. The part most people mean by “an AI feature”.
02

Agent harnesses

Routing, retries, rate limits and permissions. The layer that decides whether an agent is a demo or a system you can depend on.
03

Evaluation and observability

Test suites, tracing and scoring from week one, rather than after the first incident in production.
04

Retrieval and data pipelines

Hybrid search, reranking and chunking that holds up on documents you did not get to choose.
05

Integration into live products

Into codebases already in production, with their tests, their traces and their deployment story.
06

Backend engineering

Python and FastAPI services, queues, workers and containers. The unglamorous half that decides whether any of it runs.
Track record

What the work produced

Figures from work delivered at Kodexo Labs, Bleed AI and Appedology, before IREVEX. No customer count, no uptime number and no benchmark for Beyond Queries yet, because none of those are real yet.

75 → 90%

Query accuracy on a plain-English interface over 207 tables, after planning and schema pruning went in.

60 → 95%

Coaching agent accuracy, after an open-ended loop was rebuilt as an explicit state machine.

50%

Inference cost removed by self-hosting open models on vLLM instead of a cloud API.

35 → 10

Minutes, for a chatbot ingestion pipeline on one dataset.

Try the product, or bring us a project

Ask Beyond Queries a question about your own data, or tell us about the AI system you need built. Either way, you talk to an engineer.

Try Beyond QueriesBook a call