About SimpliSafe
SimpliSafe is a leading innovator in the home security industry, dedicated to making every home a safe home. With a mission to provide accessible and comprehensive security solutions, we design and build user-centric products that empower individuals and families to protect what matters most.
We believe in a collaborative and agile environment where learning and growth are continuous. Our teams are composed of talented individuals who are passionate about technology, security, and delivering exceptional customer experiences.
We’re embracing a hybrid work model that enables our teams to split their time between office and home. Hybrid for us means we expect our teams to come together in our state-of-the-art office on two core days, typically Tuesday, Wednesday, or Thursday – working together in person and choosing where they work for the remainder of the week. We all benefit from flexibility and get to use the best of both worlds to get our work done.
About the Role
We are seeking a highly motivated and experienced Computer Vision Applied Research Engineer to join our growing Edge AI team. As a key contributor, you will lead development of on-device machine learning for outdoor monitoring in the home security space. You will build and optimize computer vision models that run in real time on resource-constrained embedded devices like outdoor cameras and doorbell cameras, balancing accuracy with latency, memory, power, and reliability in challenging conditions (night, weather, motion blur, occlusions).
Responsibilities
- Lead end-to-end development of edge ML models for outdoor monitoring (e.g., person/vehicle/package detection, classification, tracking, segmentation, event understanding).
- Architect, train, and deploy transformer-based vision models (e.g., compact ViTs, hierarchical transformers, DETR-style detectors) and hybrid CNN-transformer backbones optimized for embedded inference.
- Drive model efficiency through resource-aware design and training, including:
- Architecture: Token/patch reduction, efficient attention variants, early-exit / conditional compute
- Training: distillation from large transformer teachers to edge students
- Compression: Quantization (PTQ/QAT), pruning, mixed precision, and operator-aware optimization
- Translate product requirements into model targets (accuracy, FPS, memory footprint, power/thermal) and ensure models meet budgets on doorbell/outdoor camera hardware.
- Partner with embedded/firmware and platform teams to integrate models into production pipelines; profile bottlenecks and improve end-to-end runtime performance.
- Define evaluation strategies tailored to outdoor edge deployments; perform failure analysis and improve long-tail robustness (nighttime, rain/snow, backlight, fast motion).
- Set technical direction and raise engineering standards: best practices for experimentation, reproducibility, model/version management, and deployment readiness; mentor other ML engineers.
Qualifications
- 8+ years in applied ML/ML engineering, including shipping production CV models.
- Strong computer vision background with deep learning expertise across detection/classification/segmentation/tracking.
- Hands-on experience with vision transformers and/or DETR-style architectures, including practical knowledge of efficiency trade-offs for edge deployment.
- Demonstrated success deploying models in resource-constrained, real-time environments (embedded/mobile/IoT/edge).
- Deep experience in model optimization: QAT/PTQ, distillation, pruning, compression, mixed precision, and hardware/runtime-aware training.
- Proficiency in Python and PyTorch and/or TensorFlow; ability to productionize models and collaborate with systems engineers (C++ experience strongly preferred).
- Staff-level leadership: ability to drive ambiguous initiatives, align stakeholders, and mentor engineers.
Bonus Points
- Expertise in efficient transformer techniques (e.g., attention approximations, windowed/local attention, KV caching where applicable, token merging, sparsity) and their deployment implications.
- Experience building model “ladders” across multiple chipsets/device tiers with consistent KPIs and automated regression testing.
- Experience with embedded inference tooling and runtimes (e.g., TFLite, ONNX Runtime, TensorRT) and model export/compatibility constraints.
- Familiarity with embedded accelerators and profiling (ARM NEON, DSP/NPU toolchains), kernel/operator tuning, and real-time video pipelines.
- Experience with long-tail data strategies (active learning, hard-negative mining) and edge reliability/telemetry feedback loops.