The rapid deployment of millions of mobile sensors and smartphones has resulted in a demand for opportunistic encounter-based networking to support mobile social networking applications and proximity-based gaming. However, the success of these emerging networks is limited by the lack of effective and energy efficient neighbor discovery protocols. While probabilistic approaches perform well for the average case, they exhibit long tails resulting in high upper bounds on neighbor discovery time. Recent deterministic protocols, which allow nodes to wake up at specific timeslots according to a particular pattern, improve on the worst case bound, but do so by sacrificing average case performance. In response to these limitations, we have designed Searchlight, a highly effective asynchronous discovery protocol that is built on three basic ideas. First, it leverages the constant offset between periodic awake slots to design a simple probing-based approach to ensure discovery. Second, it allows awake slots to cover larger sections of time, which ultimately reduces total awake time drastically. Finally, Searchlight has the option to employ probabilistic techniques with its deterministic approach that can considerably improve its performance in the average case when all nodes have the same duty cycle. We validate Searchlight through analysis and real-world experiments on smartphones that show considerable improvement (up to 50%) in worst-case discovery latency over existing approaches in almost all cases, irrespective of duty cycle symmetry.