The problem shown on this slide is that the signal-to-noise ratio (SNR) is too small. The solution procedure in this instance involves going back and reviewing the criteria selection for the diode. Perhaps there is a diode with better noise figure. The use of a more sensitive diode or a lens to magnify the light, can allow reduction of the  feedback resistor value. That will reduce all four of the components that contribute to output noise. This is where a high-quality amplifier can help the most, in addition to low bias current and offset voltage, look for an amp with low input equivalent voltage noise and low input equivalent current noise. Noise is a direct function of the square root of the bandwidth. If at all possible reduce the bandwidth of the design. Just increasing CF may not minimize noise. It will lower the signal bandwidth of the circuit, but it actually increases the closed-loop bandwidth and the noise terms that appear over closed-loop bandwidth, such as op-amp noise. It may be better to insert a filter circuit after the photodiode amplifier, which might allow reduction of the noise by a factor of ten, while keeping the same signal bandwidth. Measurement of DC light, like a photographer's light meter, requires a low-speed amplifier like the AD795, or the design can include a post-filter circuit. If the noise is still too high, consider chopping the light with a wheel and AC couple the signal path. That will take out the low-frequency noise, but the light amplitude has to be synchronously detected, by only sampling the output at the pulse tops, or by using a lock-in amplifier circuit that restores the DC level.