We consider the problem of traffic density estimation with sparse measurements from stationary roadside sensors. Our approach uses Fourier Neural Operators to learn macroscopic traffic flow dynamics from microscopic-level simulations. During inference, the operator functions as an open-loop predictor of traffic evolution. To close the loop, we couple the operator with a corrector module that combines the predicted density with sparse measurements from the sensors. Simulations indicate that, compared to the open-loop predictor, the proposed observer exhibit classical closed-loop properties such as robustness to noise and ultimate boundedness of the error. This shows the advantages of combining learned physics with real-time corrections, and opens avenues for efficient, accurate and interpretable data-driven observers.