From Home Embedded Machine Learning Jobs in Miami, FL

Position Summary
We are building automated asset identification and serialization into our equipment — vision systems that read, decode, and verify identifiers on media and drives as they move through the destruction process, and that produce the auditable records our clients rely on for chain-of-custody compliance.

This role owns that capability. You will develop and productionize computer vision and machine learning pipelines for label and barcode recognition, integrate them with machine controls and back-office systems, and serve as the internal technical owner for external engineering vendors contributing to these programs.

This is a hands-on individual contributor role with real ownership. You will write production code, and you will also be the person who defines scope for outside contractors, reviews what they deliver, and decides whether it is good enough to ship.

Key Responsibilities

Computer Vision & Machine Learning

•      Develop, train, and deploy object detection and recognition models for barcode, 2D code, and text identification on industrial asset labels.

•      Implement rotation-tolerant detection using oriented bounding box or equivalent approaches; labels arrive at arbitrary orientations, skew, and standoff distances, and the pipeline must handle it without operator intervention.

•      Solve identifier disambiguation: asset labels routinely carry multiple codes — part numbers, model codes, lot codes, vendor identifiers — with no universal marking standard. Build the detection, classification, and confidence-scoring logic that determines which code is the correct serial identifier, and define the fallback path when confidence is insufficient.

•      Integrate OCR for identifiers that are printed as human-readable text with no machine-readable code present.

•      Build and maintain the data pipeline behind the models: capture, annotation standards, augmentation, synthetic data generation, versioned datasets, and reproducible training runs.

•      Define and track model evaluation criteria — precision, recall, mAP, per-class failure analysis — and drive measurable improvement release over release.

Embedded & Systems Integration

•      Deploy inference to edge compute on production equipment; optimize for latency and throughput within the constraints of the hardware (quantization, ONNX/TensorRT, batching).

•      Specify and integrate machine vision hardware: industrial cameras, optics, lighting, triggering, and mounting geometry.

•      Interface vision output with machine control systems (PLC/HMI, serial, Modbus, Ethernet/IP) to gate machine behavior on identification results.

•      Design and build RESTful APIs and services that move serialization data from the machine into back-office and ERP systems, producing the records that support certificates of destruction and audit trails.

•      Own reliability in the field: logging, telemetry, error handling, remote diagnostics, and update mechanisms for deployed systems.

Vendor & Program Ownership

•      Serve as internal technical owner for external engineering vendors and offshore contractors working on vision, firmware, and robotics programs.

•      Write technical scopes of work, define acceptance criteria and milestone gates, and evaluate deliverables against them.

•      Review vendor code and architecture; identify where outsourced work is accruing technical debt and correct course.

•      Run technical cadence with distributed vendor teams across time zones and maintain the internal documentation that keeps knowledge in-house.

•      Support adjacent automation programs, including vision-guided robotic material handling.

Required Qualifications

•      4+ years developing production software in embedded, machine vision, or applied ML roles.

•      Strong Python; working proficiency in C or C++.

•      Practical computer vision experience with OpenCV and modern detection architectures (YOLO family or equivalent), including training and fine-tuning your own models rather than only consuming pre-trained ones.

•      Experience with PyTorch or TensorFlow, including dataset construction, training, evaluation, and deployment.

•      Barcode and 2D code decoding experience (Code 128, Code 39, Data Matrix, QR) using libraries such as ZBar, ZXing, or libdmtx — including the cases where they fail.

•      Experience deploying models to edge or embedded compute (NVIDIA Jetson or similar).

•      RESTful API design and integration; JSON, authentication, error handling.

•      Git, code review, and CI practices.

•      Ability to work onsite in Miramar, FL and directly with equipment on the production floor.

Preferred Qualifications

•      OCR experience (Tesseract, PaddleOCR, TrOCR) applied to low-quality industrial imagery.

•      Oriented / rotated object detection specifically (OBB heads, DOTA-style datasets, or equivalent production experience).

•      Industrial camera integration — GigE Vision, USB3 Vision — and practical lens and lighting selection.

•      PLC/controls integration experience in a manufacturing environment.

•      Firmware experience on ARM/STM32 and RTOS environments.

•      Vision-guided robotics: hand-eye calibration, pick-and-place, 3D or depth sensing.

•      ERP or MES integration experience (NetSuite a plus).

•      Experience managing or technically overseeing external development contractors.

•      BS in Computer Science, Computer Engineering, Electrical Engineering, or equivalent practical experience.

At Phiston Technologies, Inc., we believe in innovation, proactive product development, and secure destruction of data. We pride ourselves on being the leading physical data destruction company in the United States. Our clients include some of the top tech companies globally, such as Amazon, Facebook, and Salesforce.

Our products were engineered in response to the rapid growth of information technology and the inherent rise of data breaches across multiple industries. We deeply understand the standards that government and private entities must follow to protect sensitive information from data security breaches.

Our product line-up shows our continued commitment to ever-changing technology and data security demands. We have the only hassle-free automated crushing systems widely reviewed as a necessity in large organizations. Phiston carries a wide array of products from portable desktop destruction destroyers to enterprise-level data center destruction machines.