Senior AI/ML Engineer

Knightscope•Sunnyvale, CA

1d•Onsite

About The Position

Knightscope is seeking two Senior AI/ML Engineers to own the machine learning detection pipelines running on the Intelligent Control Module across their new K1, H1, and K7 autonomous security robots. The Intelligent Control Module runs a full edge inference stack on NVIDIA Jetson hardware, featuring a Deep Stream-based multi-model detection pipeline for people, vehicle, license plate, and face detection, all executing concurrently at real-time frame rates on constrained onboard hardware. In addition to owning the onboard detection pipeline, these engineers will also architect the AI intelligence layer for the Signals platform. This layer includes a prioritization engine, pattern detection system, recommendation scorer, explainability module, and a continuous feedback loop that transforms raw robot detections into actionable security intelligence. This is a hands-on production engineering role where you will own model training, optimization, deployment, and the ML Ops lifecycle end-to-end.

Requirements

5–10 years of software engineering experience with a focus on applied machine learning and computer vision in production environments.
Deep hands-on expertise with NVIDIA Deep Stream SDK: multi-model pipeline design, Gst-nvinfer plugin configuration, primary and secondary inference graphs, and custom output layer parsers.
Strong proficiency with YOLO-family models (YOLOv8, YOLOv9, YOLO11): training, fine-tuning on custom datasets, ONNX export, and Tensor RT engine optimization.
Hands-on experience with NVIDIA Jetson platforms (Orin NX, Xavier, or equivalent): Tensor RT INT8/FP16 quantization, DLA offloading, GPU memory management, and latency benchmarking.
Proficiency in Python for ML engineering and C++ for performance-critical inference code and Deep Stream custom plugins.
Experience building ML Ops pipelines: MLflow or equivalent for experiment tracking and model versioning, automated training with Kubeflow or similar, and production drift monitoring.
Familiarity with foundation model APIs (OpenAI, Anthropic, or equivalent) and RAG/agentic architectures for intelligence enrichment use cases.
BS/MS in Computer Science, Electrical Engineering, or related field — or equivalent professional experience.

Nice To Haves

Experience with multi-modal sensor fusion and multi-camera detection pipelines is a strong differentiator.

Responsibilities

Own and maintain the onboard detection pipeline running on the Intelligent Control Module across the new K1, H1, and K7 robots, including the Deep Stream multi-model architecture, YOLOv9/YOLO-family detection models for people, vehicle, license plate, and face detection, and GPU-accelerated inference on NVIDIA Jetson Orin NX and Xavier.
Optimize edge inference performance through model quantization (INT8/FP16), Tensor RT engine compilation, DLA offloading, and latency profiling to meet real-time frame rate targets under concurrent multi-model load.
Architect and build the Signals AI intelligence layer, encompassing a prioritization engine, pattern detection, recommendation scorer, explainability module, and human-in-the-loop feedback pipeline.
Integrate foundation model APIs (OpenAI, Anthropic, or equivalent) into the Signals intelligence stack for context enrichment, anomaly summarization, and operator-facing recommendations.
Build and maintain ML Ops infrastructure, including model versioning with MLflow or equivalent, automated training pipelines, CI/CD for model deployment, and production monitoring for accuracy drift and inference latency.
Define and maintain model evaluation frameworks, benchmark datasets, and performance regression tests to ensure detection quality across firmware and hardware updates.
Collaborate with the Intelligent Control Module Principal Architect, Full Stack engineers, and the Senior Audio/Video team to integrate ML outputs cleanly into the broader Intelligent Control Module and Signals platform.
Mentor junior engineers and contribute to architecture reviews and technical documentation for the ML stack.