NiFe LDH based materials are promising candidates for cheap, high performance catalysts for alkaline OER. While suffering from degradation effects, the presence of Fe, even in trace amounts, increases their OER activity manifold compared to plain Ni hydroxide. Although discovered in the mid-80s, the exact catalytic mechanism of this “Fe effect” is still not fully understood. Additionally, other doping elements like Co, Cr, Al and others have shown to both further improve OER activity and increase stability. The properties of those materials are often highly dependent on the doping ratio, resulting in an enormous number of possible material combinations. Modern techniques like DFT simulations and machine learning can help to speed up the search for the optimal compositions, but ultimately manual experimentation often limits how fast better materials can be discovered, as the process remains both resource and time-consuming.In this work, we combine an autonomous high-throughput experiment with machine learning assisted optimization of material compositions to a closed-loop system in an attempt to substantially accelerate the search for high performance alkaline OER catalysts. The system screens an eight variable phase space of doped NiFe LDH for compositions that yield optimal OER performance. Its centerpiece is a 4-axis robotic arm from North Robotics with sample, vial and liquid handling capabilities. The samples are prepared by means of a dip-coating process on Ni foam substrates that allows for doping with different metals. Electrochemical investigations are carried out in 1M KOH in an automated test cell, including EIS, CV and CP measurements at current densities of up to 200mA/cm2. From these results, a figure of merit for the OER activity is calculated and fed into a bayesian optimizer (Dragonfly). This algorithm determines the most promising next test candidate based on the principles of exploration and exploitation and initiates the next experiment accordingly. The experimental procedure can be divided into independent sub-steps, allowing for smart scheduling. This can reduce downtime of the robotic arm and therefore increase the throughput significantly, such that up to 1000 experiments per week can be achieved.We present the first results of a screening study conducted with this system and compare it to a manual study in terms of accuracy, reproducibility and throughput. Selected, particularly high performing compositions found in this study are also tested in a flow cell setup under industrially relevant conditions (30 wt. % KOH, 60 °C) for several hundred hours.