Engineering careers on enterprise-critical data and AI systems
Our platforms operate inside banks, ministries, hospitals and industrial plants, in Arabic and English, under active regulatory oversight. Engineers here build systems that carry real operational and compliance accountability.
- People
- 200+People
- Enterprise customers
- 200+Enterprise customers
- AI-handled interactions a year
- 600M+AI-handled interactions a year
What distinguishes the work here
- 01
Production systems, national scale
Our systems operate inside banks, ministries, hospitals and industrial plants. Work delivered here reaches production.
- 02
Broad ownership, not ticket queues
Small teams, wide ownership. Engineers talk to the client and defend their own design decisions.
- 03
Arabic AI is a specialism, not a translation task
Few organisations operate Arabic-language AI systems at this volume, which makes the experience gained here difficult to acquire elsewhere.
- 04
Teaching is part of the job
Knowledge Green cohorts are taught by the same specialists who deliver client work, with teaching time recognised as professional development.
Lakehouse platforms for national-scale programmes
Contract-first ingestion, lineage you can audit, quality gates that block a bad release. Snowflake, Databricks, Azure and open table formats, chosen per client, not per vendor relationship.
Arabic NLP validated on real customer speech
Dialect handling across Gulf, Levantine and Egyptian speech, Arabic entity resolution, RTL product surfaces, and evaluation sets written in Arabic instead of translated from English benchmarks.
Agents that execute actions in core systems
Tool-calling systems wired into core banking, claims, ERP and ticketing, with permissioning, human gates, replayable traces and a rollback path for every action an agent can take.
In-Kingdom and in-boundary deployment
Running modern AI where the data is not allowed to leave: private inference, in-country hosting, network isolation, and evidence packs that satisfy NDMO and PDPL reviewers.
Routing across a model portfolio
Cost-to-quality engineering: smaller models handle the majority of workloads, with larger models applied only where evaluation evidence justifies them and automated evaluations governing the routing.
Time series, digital twins and operational technology
Sensor pipelines, anomaly detection and digital twins for plants and networks where a false positive stops a line and a false negative breaks one.
The systems you would work on
Technical surfaces of substantial engineering complexity, each running in production for enterprise clients under confidentiality agreements.
The disciplines you would join
We recruit against these families throughout the year rather than by individual advert. Each family sets out the scope, the first 90 days and the performance standard applied.
Scope
Own the ingestion and modelling layer of a client platform end to end: sources, contracts, transformation, quality gates and the cost of running it.
First 90 days
Replace a fragile nightly batch with an incremental, tested pipeline; put lineage and freshness in front of the business owner for the first time.
The bar
Able to interpret a query plan, justify a modelling decision and diagnose pipeline performance from evidence.
If your discipline is not listed, we still welcome your application; new practice areas are frequently created around exceptional candidates. mt@bilytica.com
The selection process, typically within two weeks
Candidates receive a clear status at every stage, and every application is reviewed by a member of the delivery team.
Application reviewed by a delivery specialist
Within 5 working daysApplications are reviewed by someone who performs the role, and every candidate receives a response.
Conversation about your work
45 minutesA 45-minute discussion of work you have delivered, including what failed and what you would approach differently.
Working session, paid if it takes real time
90 minutesA real problem from a redacted engagement, worked through together in 90 minutes. No unpaid multi-day take-home.
Meet the partner for your practice
30 minutesEvery hire meets a founding partner, giving candidates direct access to the leadership of the practice they are joining.
Offer with the scope written down
Within 2 weeks end to endThe offer specifies the practice, the first engagement and the reporting line, so expectations are clear from day one.
Teaching is part of the role
A defined share of Bilytica revenue funds free data and AI education in the markets where we operate, delivered by our own engineering and consulting teams. Employees are allocated paid time to lead cohorts, publish frameworks and mentor students as part of their professional development.
See the programmesSend work, not a template
A repository, a design document or a post-incident review, any material that demonstrates your approach. Include the role family in the subject line and a member of the team will respond within five working days.
Careers: FAQ
Roles, locations, the hiring loop and teaching time.
Does Bilytica have open roles right now?
Bilytica hires continuously against evergreen role families, data engineering, analytics engineering, AI/ML engineering, Arabic NLP, agent engineering, solution architecture and delivery consulting, rather than posting one-off adverts. Applications for any of these families are read year-round.
Where can I work from?
Bilytica has teams in Riyadh, Dubai, Melbourne, Madrid, Lahore and Finland. Most engineering roles are based in Riyadh, Dubai or Lahore and are remote-friendly within a compatible time zone; delivery-facing roles need regular presence with the client.
What does the hiring process look like?
A structured process, usually completed within two weeks: a delivery specialist reviews the application and responds within five working days, a 45-minute discussion of delivered work, a 90-minute working session on a redacted client problem, a 30-minute meeting with the founding partner for the relevant practice, and a written offer specifying practice and first engagement.
Is there an unpaid take-home assignment?
No. Bilytica does not use multi-day unpaid take-home tests. The technical stage is a 90-minute working session done together, and is paid if it runs beyond that.
Do engineers get to teach?
Yes. Bilytica's Knowledge Green programme runs free cohorts for students, teachers and public-sector teams, taught by the same specialists who deliver client work. Teaching time is allocated within working hours as part of the role.