About Think
We make the world’s compute radically more efficient.
Think AI was founded in Riyadh in 2025 on one problem: expensive silicon sitting underused, held back by software blind to the hardware
it runs on. We design and build the machine and the orchestration together, in one building, so that stops being true.
Our Mission
To redefine AI infrastructure from the silicon up.
To build the most efficient compute platform on earth, where hardware and software co-evolve to maximise intelligence. To ignite a new
golden age of engineering, where anyone can own, operate, and scale frontier AI without compromise, without dependency, and without
wasted compute.
Our Values
Four principles run through everything we do. We look for these in how people demonstrate the values in their everyday work, not in how
they talk about themselves.
Ownership مسؤولية
You own the outcome, not just the task. You’re frugal with resources and you put the mission ahead of your own comfort.
Agility مرونة
You move with velocity. You ship quickly and reiterate and once a decision is made, you disagree and commit.
Impact أثر
You build for the customer. You think big and execute simply and you measure yourself on outcomes, not activity.
Mastery إتقان
You go deep on the hard problem. Excellence is your default and you take real pride in delighting the customer.
About the role
You will own the software roadmap end to end across Full Stack, Web, Machine Learning, DevOps and Site Reliability five disciplines that today are separate teams and need to behave like one system. Think Fabric and ILM are the products. Everything else serves them.
You are the one who sets the bar for what counts as a result, builds the organisation that clears it and hires the team who run each discipline under you.
You will partner with Hardware Engineering on the telemetry interface between the Node and ILM.
What you will do
Own the software roadmap across Full Stack, Web, ML, DevOps and SRE, sequenced against the hardware roadmap rather than alongside it.
Own ILM's serving path: batching, memory management, cache behaviour and scheduling under real concurrency and the architectural decisions that govern them.
Set the engineering bar. Design review, benchmark standards, release gates, production readiness. If it is not measured on real hardware it is not a result, and you are the one who enforces that
.Make models share accelerators well, including across silicon of different capacities and generations. This is the core commercial claim of the platform.
Own reliability. Uptime, incident response, postmortem discipline, and the corrective-action culture around it.
Build the organisation: grow a multidisciplinary team, hire the leaders under you, and set the levelling and performance standard for engineering.
Partner with Hardware Engineering on the telemetry interface between the Node and ILM, and on the co-design decisions that only work if both roadmaps move together.
Be in the code enough to be credible. Read a profile, review a benchmark, call a design decision wrong with evidence.
What you need
Ten years or more in software engineering, including four leading engineering teams with direct experience hiring and managing managers, not only individual contributors.
Depth in at least two of: ML inference or training infrastructure, distributed systems at scale, cloud or platform infrastructure, or high-performance computing.
You have taken software from architecture through production at scale and carried the operational consequences not just shipped a launch.
Hands-on fluency that survives scrutiny. Strong Python and a systems language, typically C++ or Rust. You have personally optimised inference or training throughput and can explain exactly where the win came from.
Understanding of accelerator memory hierarchies, kernel launch behaviour and where time actually goes. You can tell a real bottleneck from an assumed one.
Comfort on bare metal rather than behind a managed service. We own the hardware; the abstractions are ours to build.
Production reliability ownership: on-call structure, SLOs, incident command, and the discipline to make postmortems change behaviour.
A player-coach. In the design review, in the incident channel, and at the customer site when the architecture is the argument.
Based in Riyadh or willing to relocate.
Useful, not required
CUDA, ROCm, Triton or comparable kernel-level work.
Contributions to vLLM, TensorRT-LLM, SGLang or similar open-source serving stacks.
Distributed serving and multi-accelerator sharding at production scale.
Published benchmark or systems work.
Experience in a company where hardware and software teams shipped a co-designed product.
Founding engineer or engineering leadership at a deep-tech company that reached commercial scale.
Arabic language proficiency.
What we offer
One of the first deep tech companies in the region, building foundational technology in house.
Meaningful ownership and impact at an early stage.
Competitive early-stage compensation.
Close collaboration with a small, senior team.
Problems that combine hardware, systems and AI at scale.
