Computer Vision & ML Mentor

This is a part-time mentor-first role. You will help two strong full-stack AI engineers move faster and make better calls on the hard computer vision and machine learning problems and you will help us put honest measurement around what we ship. You are not here to own delivery or write the bulk of the code. You are here to raise the bar unblock the difficult problems and make sure the model work is sound and measurable before it reaches a customers floor.

• Review the detection and vision-language pipeline and the architecture choices and steer the team toward approaches that hold up across several live sites.
• Set up honest evaluation: how we measure alert precision missed events and time to detect per site so a rule that cannot be measured does not ship.
• Coach the two engineers through the hard problems (model selection data and labeling false-positive control edge versus cloud trade-offs) in regular working sessions and design or code reviews.
• Help us avoid expensive mistakes early: over-engineering alarm fatigue and GPU cost surprises.
• Sanity-check our edge and NVIDIA plans (Jetson TensorRT DeepStream) against what we can realistically build and run at this stage.
• Be a steady technical sounding board for the founder on what is proven what is risky and what to do next.

• Senior hands-on experience in computer vision and applied machine learning ideally with real production deployments. Live video or camera work is a strong plus.
• A track record of mentoring or leading engineers and the patience to teach rather than take over.
• Strong instincts for evaluation and measurement not only for building models.
• Comfort with a small early team and a pragmatic ship-and-measure mindset.
• Clear communication in English. Arabic is a bonus not a requirement.
• A few reliable hours a week and the willingness to be reachable when the team is genuinely stuck.

Nice to have

• Experience with vision-language models edge inference or real-time video pipelines.
• Familiarity with the NVIDIA stack (Jetson TensorRT DeepStream Triton).
• Industrial manufacturing or safety-monitoring context.
• Experience advising or working inside early-stage startups.

• A light well-scoped commitment that fits around your main work: advisory equity a part-time retainer sized to the hours or a mix.
• Direct influence on a real Arabic-first AI product at the stage where good guidance matters most.
• A small honest team that will use your input and tell you straight what worked.
• A clear scope so the time stays predictable and respected.