Designed and implemented a modular enemy AI system featuring multiple enemy archetypes with distinct behaviors and combat roles. This included heavy tank enemies that move slowly but can absorb significant damage, as well as fast, high-damage enemies with low health that pressure the player through aggressive tactics. Enemies use both sight and hearing to detect the player, reacting to gunfire and footsteps to create more dynamic and believable encounters. A memory system allows enemies to lose track of the player when line of sight or audio cues are broken, preventing artificial omniscience. To improve performance, I implemented a proximity-based activation system using a trigger sphere around the player, ensuring only relevant enemies are active and updating at any given time.