With 1. d4 d5 2. c4 dxc4 3. Nf3 a6 4. e3 b5, i won't burden with the back-story, but long story short, I'm still tinkering with my repertoire before the Reykjavik open.

I've analyzed this out as though it were a centaur game, drilling into variations with SF 17 to depth 50+, and this is what I've come up with as best play, though of course there are sidelines that would need to be accounted for.

1. d4 d5 2. c4 dxc4 3. Nf3 a6 4. e3 b5  5. a4 Bb7 6. b3 Nd7 7. bxc4 bxc4 8. Bxc4 e6 9. Nbd2 Ngf6 10. O-O c5 11. Rb1 Qc7 12. Ba3 Be7 13. Bd3 O-O 14. Qc2

The TN here I guess is 7...bxc4 rather than 7...b4?!? as in Dubov-Popov, Tyumen, RUS, 2012

Then engine gives just a 0.1 to 0.25 eval (depth 52) dis-advantage, nevertheless seems quite sharp to me, so some winning chances.  Plus it would be a surprise rather than just the same-old same-old compared with 3...Nf6