We consider a risk averse entrepreneur who approaches a diversified venture capitalist (VC) for financing of a project with positive potential return. We develop several models that capture key features of the venture financing, including staged investment, VC oversight costs and agency conflicts. The contract between the VC and the EN includes risk-free and pay-performance sensitive compensation. Moral hazard arises because the EN must exert effort for the project to succeed. Our model is novel in that it also allows for asymmetric beliefs about project quality due to the EN's optimism even when the VC and EN face symmetric information. We first analyze the VC-EN relationship when the VC has bargaining power. We characterize the equilibrium levels for the pay-performance sensitivities, investment and effort over time and show they can be either increasing or decreasing or initially increasing and then decreasing. We find that asymmetric beliefs and risk aversion have opposite effects on the VC-EN relationship. When the EN is moderately more optimistic than the VC, he accepts more risk and exerts more effort and the VC responds with more investment. In contrast, risk aversion reduces effort and investment. Our model predicts a performance-sensitive investment policy where critical milestones must be achieved for investment to continue. These milestones increase with the risk aversion and decrease with the asymmetry in beliefs. Consequently, project duration increases with asymmetric beliefs and decreases with risk aversion. We calibrate this core model to empirical data and use numerical analysis to demonstrate that the technical and systematic risks have opposite effects. The VC's payoff and the project's value and duration increase with technical risk and decrease with systematic risk. We analyze the relationship when the EN has bargaining power, and find that the equilibrium and the corresponding implications for venture financing do change. In this setting, the negative effects due to risk aversion are more pronounced. We also find that if the EN's effort cannot be observed by the VC, then the pay-performance sensitivities, investment and effort all increase.