The local dark matter density is actually quite tiny, on the order of $\rho\sim10^{-19}\text{ g/cm}^3$ (see e.g. Bovy & Tremaine (2012)). This means that there is roughly $0.001$-$0.01M_{\odot}$ of dark matter per cubic parsec - a staggeringly small amount. 1000 cubic parsecs would contain about one solar mass of dark matter - and that's a cube 10 parsecs in length on each side! Now, the distribution of dark matter in galaxies is not homogeneous - it follows, roughly, a Navarro-Frenk-White profile, decreasing in density from the center of the galaxy - but on the scale of parsecs (and certainly in the Solar System), we can consider it to have roughly uniform density.
On small scales, then, we have approximate homogeneity and low density. This means that any gravitational lensing effects from dark matter should be extremely low or self-cancelling, arising only from inhomogeneities containing large clumps of dark matter. However, such clumps are unlikely to form solely through dark matter's interaction with itself (if we discount the MACHO hypothesis, which, as far as I know, is not currently favored).
On intergalactic scales, however, dark matter can have some effects. Weak lensing is a commonly-observed phenomenon in galaxy clusters, which may have extremely high fractions of dark matter. There are several techniques currently used to model the mass distribution of the lensing galaxy (see the KSB+ method) and to reconstruct the image and position of the original galaxy via deconvolution (see Chantry & Magain; a visual example is given here). I'm not too familiar with either technique, though, so I can't give you a good overview.
Even large-scale lensing has large mass requirements. zephyr pointed out that the foreground object that created the Einstein cross contained $\sim10^{10}M_{\odot}$ of dark matter (van de Ven et al. (2010)). That is enormous!
