I will discuss our work on the development of human Organ-on-a-Chip (Organ Chip) microfluidic culture devices lined by living human cells to replace animal testing, accelerate drug development, discover new biomarkers, and advance personalized medicine. Our Organ Chips are effectively living 3D cross sections of major functional units of living organs that contain human organ-specific epithelial cells interfaced with human microvascular endothelial cells that are exposed to fluid flow to mimic vascular perfusion. We also recreate the relevant physicochemical microenvironment of each organ, for example, by recreating breathing motions and an air-liquid interface in lung and trickling flow and peristalsis-like deformations in intestine. We have engineered multiple human Organ Chips, including lung (alveolus and airway), intestine (duodenum, ileum, colon), kidney (proximal tubule and glomerulus), bone marrow, liver, and blood-brain barrier (BBB) chips, as well as fluidically linked BBB Chips and brain neuronal network chips. This Organ Chips have been used to develop various human disease models and uncover new drug targets and potential clinical biomarkers, as well as discover new therapeutics. In addition, multiple different human Organ Chips have been fluidically linked to create an automated 'human Body-on-Chips' for real-time analysis of cellular responses to pharmaceuticals, chemicals, and toxins, as well as for quantitative in vitro-to-in vivo translation (IVIVT) of human drug pharmacokinetics in vitro.

In machine learning, we use data to automatically find dependences in the world, with the goal of predicting future observations. Most machine learning methods build on statistics, but one can also try to go beyond this, assaying causal structures underlying statistical dependences. Can such causal knowledge help prediction in machine learning tasks? We argue that this is indeed the case, due to the fact that causal models are more robust to changes that occur in real world datasets. We discuss implications of causal models for machine learning tasks, focusing on an assumption of 'independent mechanisms', and discuss an application in the field of exoplanet discovery.

The 300-meter Arecibo telescope is one of the technical wonders of the 20thcentury and had been the world leader in absolute sensitivity in deca-centimeter bands for more than half a century, until Sep. 25, 2016, the dedication of the Five-hundred-meter Aperture Spherical radio Telescope (FAST), which has two to three times the sensitivity and one-order-of-magnitude higher survey speed. Still under commissioning, FAST (Nan et al. 2011) has discovered >60 new pulsars and realized an unprecedented commensal-survey mode (Li et al. 2018). We strive to take the next step toward space through innovative radio technologies.

Diamond materials in different forms, including single crystal, polycrystalline, or nanoparticles, feature a number of very attractive physical/chemical properties for the development of both chemical or biochemical sensors. In the last decade considerable work has been carried out to process diamond (seeding, etching, patterning, controlling surface chemistry, etc.) so that it has now become a serious candidate material for implementation in advanced sensor technologies. Here we report on some of the work achieved by our team in this area, focusing more particularly on sensors used for the detection of small organic compounds both in the liquid phase or gas phase.