Fluorescent proteins (FPs) have revolutionized our ability to image biological systems. However, poor penetration and high autofluorescence of visible light in mammalian tissue has motivated the design of FPs that can take advantage of the minimal autofluorescence, scattering, and absorption in the near-infrared (NIR) “optical window” (650-900nm). Genetically-encoded NIR-FPs have been designed from various classes of photoreceptors to covalently attach biliverdin (BV), a heme metabolite that is endogenous to mammalian tissue, as their chromophore. While these proteins have been engineered to have modest quantum yields and fluorescence in vitro, performance in mammalian cells does not correlate with their biophysical properties. My thesis research explores two novel strategies for developing NIR-FPs with an attempt to better understand the key determinants of their performance in mammalian cells. In our first approach detailed in Chapter 2, we developed a de novo designed fluorescence-activating protein that autocatalytically attached BV in vitro, but was plagued by poor BV binding in cells. From this experience, we hypothesized that the disparity between brightness in vitro vs in cells is caused by inefficient BV incorporation into existing scaffolds. In Chapter 3, we developed an experimental method to measure the fraction of BV-bound FP out of mammalian cells and found that chromophorylation in these scaffolds is inefficient and can help explain the disparity between cellular and molecular brightness. Through mathematical modeling, we explore the parameters that contribute to this inefficiency. Finally, in Chapter 4, we developed a new NIR-FP based off a bilin lyase protein, a novel scaffold that we hypothesized would have high chromophorylation due to its role as a chaperone protein responsible for bilin attachment in cyanobacteria. Our engineered bilin lyase binds biliverdin and fluoresces in mammalian cells, and future engineering work will focus on turning this scaffold into a viable FP.