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.
- Lead the work and the people. Own delivery end to end, mentor engineers, and partner across the company with our AI research and computer vision engineers, the wider engineering team, and the CTO, Product Managers and senior management on roadmap and direction.
Our Ideal Requirements
- Strong hands-on proficiency in Python as a primary working language.
- Strong hands-on experience with PyTorch.
- Solid production experience with DeepStream and GStreamer for building real-time video and streaming inference pipelines.
- Confident with OpenCV and image processing tooling more broadly.
- Strong foundation in data structures, algorithms and software engineering principles. You write code a team can build on and scale.
- Experience building high-throughput data ingestion pipelines or production computer vision pipelines: systems that ran at real volume, not prototypes.
- 4+ years building and deploying AI or computer vision systems in production.
- Leadership and ownership experience. A track record of owning a product or codebase end to end, driving initiatives independently, and being accountable for the outcome.
- Excellent debugging skills. You can reason your way to a root cause in a complex, concurrent, real-time system rather than changing things until it works.
- Experience deploying across both on-premises and cloud environments, and an understanding of how differently each behaves.
- Hands-on with AI-assisted development tools (Claude Code, Cursor, Copilot or similar), and an instinct for automating repetitive work.
Why Work At Ailytics?
- Lasting impact at the intersection of AI and safety. Our systems are used daily by enterprise customers across Asia and beyond, in environments where a missed detection genuinely matters.
- We're a startup still in its early stages, so there's real room to shape both the product and the way we build it.
- Full architectural freedom on a hard problem. You own the codebase and the critical decisions, and real-time inference across thousands of cameras is a genuinely difficult systems problem.
- Scale that is real today, and data you cannot get elsewhere. Thousands of live cameras running continuously in production, generating a proprietary, global-scale video dataset from real industrial environments.
- We value strong standards, high transparency, and low egos. Let us know if this sounds like you.
