Quantum noise is the key challenge in Noisy Intermediate-Scale Quantum (NISQ) computers. Limited research efforts have explored a higher level of optimization by making the quantum circuit resilient to noise. We propose and experimentally implement QuantumNAS, the first comprehensive framework for noise-adaptive co-search of variational circuit and qubit mapping. Variational quantum circuits are a promising approach for constructing quantum neural networks for machine learning and variational ansatzes for quantum simulation. However, finding the best variational circuit and its optimal parameters is challenging in a high-dimensional Hilbert space. We propose to decouple the parameter training and circuit search by introducing a novel gate-sharing SuperCircuit. The SuperCircuit is trained by sampling and updating the SubCircuits in it and provides an accurate estimation of SubCircuit performance trained from scratch. Then we perform an evolutionary co-search of SubCircuit and its qubit mapping. The SubCircuit performance is estimated with parameters inherited from SuperCircuit and simulated with real device noise models. Finally, we perform iterative gate pruning and finetuning to further remove the redundant gates in a fine-grained manner. Extensively evaluated with 12 QML and VQE benchmarks on 10 quantum computers, QuantumNAS significantly outperforms noise-unaware search, human and random baselines. For QML tasks, QuantumNAS is the first to demonstrate over 95% 2-class, 85% 4-class, and 32% 10-class classification accuracy on real quantum computers. It also achieves the lowest eigenvalue for VQE tasks on H2, H2O, LiH, CH4, BeH2 compared with UCCSD baselines. We also open-source QuantumEngine (https://github.com/mit-han-lab/pytorch-quantum) for fast training of parameterized quantum circuits to facilitate future research.