Department: Johns Hopkins Center for Bioengineering Innovation & Design (CBID), Department of Biomedical Engineering, Whiting School of Engineering

 Location & Duration: Baltimore, MD, USA (in-person job) 40 hours a week 

 Reports to: Dr. Soumyadipta Acharya (Principal Investigator)

Key Responsibilities

Lead the design, training, and evaluation of computer vision models for mosquito identification and other relevant projects in vector-borne diseases. Develop and maintain a scalable training and evaluation pipeline for image classification and detection models. Continuously explore and evaluate new architectures, training approaches, and optimization strategies to improve model accuracy and robustness. Design and maintain systems for dataset management, ensuring training, validation, and test datasets remain clean, versioned, and traceable. Maintain high standards of data organization and reproducibility across experiments and training pipelines. Develop strategies for deploying models across mobile edge devices and cloud infrastructure. Optimize models for inference on smartphones and other resource-constrained platforms. Work closely with software engineers to integrate models into the Android application and imaging pipeline. Investigate and troubleshoot performance issues related to camera systems, imaging conditions, and device variability. Develop benchmarking and evaluation methods to continuously monitor model performance across deployments. Apply statistical reasoning when evaluating model performance and clearly communicate the statistical basis for model improvements and algorithmic decisions. Collaborate with entomologists and field teams to improve data collection, labeling, and training dataset quality. Contribute to publications and presentations describing algorithm development and system performance.

Technical Focus Areas
Computer Vision and Model Development: Design and train deep learning models for insect classification and morphological recognition. Experiment with architectures such as EfficientNet, YOLO, Vision Transformers, and other modern computer vision models to determine optimal approaches for the application. Develop strategies for handling limited datasets, noisy data, and challenging real-world image conditions.

Model Optimization for Edge Deployment: Optimize models for deployment on smartphones using frameworks such as TensorFlow Lite, PyTorch Mobile, or ONNX. Investigate quantization, pruning, and other model optimization techniques to ensure efficient inference on resource-constrained devices. Ensure models perform consistently across different smartphone cameras and hardware configurations.

AI Data Pipeline and Dataset Management: Develop systems for dataset versioning, experiment tracking, and model reproducibility. Ensure that training, validation, and testing datasets are well organized, auditable, and traceable. Maintain clear documentation of dataset lineage and experiment configurations. Build workflows that support continuous model retraining as new field data becomes available.

System Architecture for AI Deployment: Design the architecture for managing model updates, versioning, and deployment across edge devices and cloud platforms. Develop strategies for monitoring model performance and maintaining reliability across large-scale field deployments.

Project Impact
VectorCam aims to transform how mosquito surveillance is conducted in malaria-endemic regions by enabling rapid and accurate species identification directly in the field. By improving the speed and accessibility of entomological surveillance, this technology has the potential to strengthen malaria control programs and support more targeted vector control interventions. This role offers the opportunity to work on a globally impactful technology while solving challenging problems at the intersection of computer vision, edge AI, and public health innovation.