Machine Learning Team Lead

Department: Engineering

Location: Taipei - Neihu

Employment Type: FullTime

Berry AI builds AI-powered operations platforms for QSR restaurants — drive-thru analytics, loss prevention, and store management tooling deployed at thousands of locations across the US — and growing. Computer vision sits at the core of what we ship: models running on in-store edge devices, reading video in real restaurant conditions to understand vehicles, people, queues, and operational events. We're hiring a Machine Learning Team Lead to own that technical direction and lead the ML/CV team behind it.

What you'll work on

• Lead and grow the ML/CV team — set technical direction, own execution quality and delivery, mentor engineers, and own hiring — while staying hands-on.

• Own the CV capabilities behind the product — object detection, multi-object tracking, person/vehicle Re-ID, camera calibration and geometric vision, and event recognition.

• Work with Product to turn customer problems into well-defined ML problems — the success metric and a roadmap the team can actually execute.

• Drive the full ML lifecycle — data collection, annotation strategy, training, evaluation, deployment, monitoring, and the iteration loop that keeps accuracy climbing after launch.

• Push performance under real-world conditions — occlusion, low light, lens distortion, camera movement, viewpoint variation, and domain shift across thousands of stores.

• Guide model optimization for constrained edge hardware — balancing accuracy against latency, throughput, memory, and operational reliability.

• Build reusable pipelines and tooling that take manual effort out of data prep, store configuration, camera calibration, validation, and rollout.

You're a strong fit if you have

• 7+ years in machine learning, deep learning, or computer vision, with a track record of shipping production CV systems — not just research prototypes.

• Experience leading an ML team, serving as technical lead, or owning the direction and delivery of a major ML product.

• Real depth in one or more of object detection, multi-object tracking, person/vehicle Re-ID, segmentation, or video understanding.

• A strong classical CV and image-processing foundation — feature extraction and matching, geometric transformations, interpolation, and image warping.

• The judgment to evaluate CV systems under real-world conditions — occlusion, truncated objects, low light, motion blur, camera variation, domain shift — and to tell which failure modes actually matter to customers.

• End-to-end ML product lifecycle experience, from dataset development and training through production deployment and monitoring.

• Python and PyTorch or TensorFlow, with a solid grasp of model evaluation, error analysis, dataset quality, and experiment design.

• The ability to turn ambiguous product needs into measurable objectives and executable plans, work across functions, and stay technically hands-on while leading the team.

Bonus points

• Camera geometry and calibration — pinhole camera models, intrinsic and extrinsic parameters.

• Edge AI and on-device inference — TensorRT, ONNX, OpenVINO, NVIDIA Jetson, DeepStream — plus quantization, pruning, distillation, or other inference optimization techniques.

• MLOps experience — dataset and model versioning, experiment tracking, automated evaluation, model monitoring, and progressive rollout.

• CV in uncontrolled physical environments — autonomous driving, robotics, security, industrial inspection, or smart retail.

• Multimodal models, VLMs, or LLM-based workflows used in real production systems.

• Startup or fast-scaling product org experience, or working with international teams and enterprise customers.

Our engineering culture

Small team, high ownership, fast feedback from customers — and the operational rigor to make that velocity sustainable. Modern AI tooling — LLMs, coding agents, agent-driven workflows — is a normal part of how we work, and you're encouraged to push on what these tools can do.

============================

Interview Process

1. Online (Google Meet)

   1. VP of Engineering Interview (1 hr)

2. Take-Home Assignmen

   1. Submission deadline: within 7 days

   2. Estimated time required: 4 hours

      Note: Deadlines can be extended upon request to accommodate current workloads or personal commitments.

3. Onsite

   1. Technical Interview (2.5 hrs)

   2. CEO & VP Interview (1.5 hrs)

   3. HR Interview (0.5 hr)