Disentangling the supply and demand factors of inflation is a complex yet crucial endeavor for effective monetary policymaking. A significant challenge arises from the fact that key elements of inflation dynamics, such as the output gap and inflation expectations, are often unobserved. This study implements a Hemisphere Neural Network whose peculiar architecture allows the estimation of the unobserved states within an augmented New Keynesian Phillips Curve. In the context of Philippine inflation, the estimated latent states effectively capture macroeconomic concepts such as real activity, inflation expectations, and commodity shocks, as evidenced by their correlation with input variables and alignment with major economic events. Long-run expectations is found to remain steady between 3.5-4.5 percent, while commodity shocks account for most spikes in realized inflation. The model’s estimated output gap is aligned with existing measures from the Bangko Sentral ng Pilipinas, while the estimated inflation expectations align with short- to medium-term expectations from businesses and professional forecasters. Finally, the research offers significant insights into inflation dynamics and provides an analytical tool for monitoring demand- and supply-side inflationary pressures.