The sixth generation (6G) systems will likely employ orthogonal frequency division multiplexing (OFDM) waveform for performing the joint task of sensing and communication. In this paper, we design an OFDM system for integrated sensing and communication (ISAC) and propose a novel approach for passive target detection in an indoor deployment using a data driven AI approach. The delay-Doppler profile (DDP) and power delay profile (PDP) is used to train the proposed AI-based detector. We analyze the detection performance of the proposed methods under line of sight (LOS) and non-line of sight (NLOS) conditions for various training strategies. We show that the proposed method provides 10 dB performance improvement over the baseline for 80% target detection under LOS conditions and the performance drops by 10-20 dB for NLOS depending on the usecase scenarios.