As part of my Master's coursework,  I worked on a class project building on CausalSim, a causal framework for unbiased trace-driven simulation. CausalSim requires Randomized Control Trials (RCTs) which are expensive. In a team of three, we investigated action invariance as a replacement to CasusalSim’s expensive RCT property through an Adaptive Bit Rate environment, and evaluated the accuracy of CausalSim models trained on this data.