Cancer is a significant global health concern; and traditional therapies, including chemotherapeutics, are often simultaneously toxic yet ineffective. There is a critical need to develop targeted cancer therapeutics which specifically inhibit molecules or molecular pathways essential for tumor growth and maintenance. Furthermore, a targeted therapy is only effective when a patient's tumor expresses the molecular target; therefore, companion diagnostics, including molecular imaging agents, are a necessary counterpart of targeted therapies. Mesothelin (MSLN) is a cell surface protein overexpressed in numerous cancers, including triple-negative breast, pancreatic, ovarian, liver, and lung, with limited expression in normal tissues. Aberrant MSLN expression promotes tumor progression, metastasis, and therapeutic resistance, and is correlated with poor prognoses. Promising results from pre-clinical and clinical trials to target MSLN with antibodies, antibody derivatives, immunotoxins, and antibody-drug conjugates for therapy demonstrate the promise of MSLN-targeting methods; however, none are currently approved for routine clinical use, and limitations are emerging for targeting agents under development. New targeted diagnostic and therapeutic approaches for MSLN-positive tumors have potential for substantial impact in the clinic. In this work, I used yeast-surface display and directed evolution to engineer novel proteins based on the non-antibody fibronectin (Fn3) scaffold that bind to MSLN with high affinity and specificity. Soluble engineered proteins were expressed and purified to high yields from a bacterial system. In vitro cytotoxicity and apoptosis assays demonstrated the potential of the proteins as targeted cancer therapeutics for MSLN-expressing tumors, and the engineered proteins enhanced cancer cell sensitivity to chemotherapy. Towards the goal of using engineered Fn3 variants for targeted drug delivery applications, in collaboration with our collaborators, we synthesized and validated a novel protein-polymer conjugate drug delivery system. Finally, I established and validated appropriate in vivo tumor models to evaluate our engineered proteins as molecular imaging diagnostic agents. My work has provided a candidate set of engineered proteins that can be used as molecular targeted therapeutics, drug-delivery vehicles, and molecular imaging diagnostics for MSLN-positive tumors, towards the ultimate goal of providing targeted treatment and diagnostic options for cancer patients where no such treatments currently exist.