You have a microcontroller in there already anyway, so have it drive the boost converter directly. Many many micros come with PWM generators intended to run switching power supplies, motor drivers, and the like. If you don't know where to start, check out the dsPIC 33 series.
Using a micro to drive a single step converter simplifies things in this case. You could use a small low side current sense resistor that develops a few 100 mV at the maximum of 700 mV output current. Let's say you use 1 Ω for sensing the current. That means the resulting voltage will be 0-700 mV. With a 12 bit A/D and 3.0 V reference, that still gives you a resolution of over 950, or almost 10 bits. That should be plenty.
Since a little bit of ripple isn't a big deal when driving LEDs, you could use a simple pulse on demand system. Every 10 µs, for example, you take a new A/D reading and decide whether the current is above or below the regulation threshold. You then either do a full size PWM pulse or not. Since the regulation threshold is in firmware, it can be easily changed via commands from the host, or a dimmer pot input, or whatever.
A more fancy scheme keeps the PWM running, but tweaks the duty cycle up and down depending on what the actual current is relative to the desired value. You could do a simple scheme where the PWM duty cycle is bumped up or down a small amount each pulse, or go all the way with a full blown servo controller. Unless a few percent ripple actually matters, the latter is overkill.
The boost converter in this case would require a low side switch, inductor, Shottky diode, and cap across the LED and resistor series string. For simplicity, use a NPN as the low side switch as it will be simpler to drive from the 3.3 V PWM digital signal. For a little more efficiency, use a N channel FET.
Take a look at my KnurdLight project. It's using a tiny micro to drive the switch of a boost converter to regulate the current thru a string of LEDs. In this case the current was fixed at 20 mA and the current sense resistor output was compared to a 600 mV reference built into the micro. This micro has no PWM generator, so the pulse are generated directly by firmware instructions. The firmware waits in a loop until the current falls below the regulation threshold, then it emits a pulse and goes back waiting for the current to be below the threshold again. And yes, it works very nicely.
