Senior AI Engineer (Remote- Within 2.5Hrs of SGT)

Here at Ailytics, we're building AI solutions to envision a safer world. By combining computer vision and predictive analytics, we enable organizations to proactively identify risks, optimize processes, and ultimately save lives. Our platforms are currently deployed all over the world, covering more than 500 million square meters!

You will own the computer vision pipeline that our entire product runs on: a real-time system processing live video from thousands of cameras simultaneously, serving our heaviest and most accurate models under hard latency constraints. Every detection a customer acts on passes through this codebase.

The engineering problem is a genuinely difficult one. Running our most accurate models in real time, at that camera count, on hardware that is often sitting on a customer's premises rather than in a datacentre, means constantly trading throughput against accuracy against cost against what the hardware can physically do. Getting those trade-offs right is the core of this job.

You will own the codebase and make the critical architecture decisions, with full freedom. We are hiring you to make those calls, not to implement someone else's. You will work directly with our CTO, Product Managers and senior management, and you will lead the engineers building alongside you.

To be clear on scope: this role is focused on production computer vision systems and real-time inference at scale. It partners closely with our generative AI and model research work, but it is a distinct role. This is where the pipelines and the systems live.

What You Will Be Doing

-Own and drive the computer vision codebase. Take full technical ownership of the pipeline, its architecture, its quality and its direction, with the freedom to make the critical calls.
-Build scalable, optimised real-time CV pipelines using Python, PyTorch, DeepStream and GStreamer, across both on-premises and cloud environments.
-Scale to thousands of cameras. Design for concurrency, throughput and graceful degradation, so adding cameras does not mean rewriting the system.
-Serve our heaviest and most accurate models in real time. Optimise inference through batching, GPU utilisation, precision and pipeline design, and know exactly where the bottleneck is at any moment.
-Build high-throughput data ingestion. Move large volumes of video and detection data reliably through the system without loss or backpressure problems.
-Benchmark and improve model performance in production conditions, refining architectures and the surrounding code for accuracy, speed and stability.
-Debug the hard problems: pipeline stalls, dropped frames, memory leaks, GPU saturation and non-deterministic failures at scale, the issues that only appear in production.-Automate everything worth automating. Write the scripts and internal tooling that make deployment, testing and diagnosis faster for the whole team, and use AI tooling day to day to accelerate your own work.