This paper presents a new multilevel hypergraph partitioning algorithm, hMEfS, which is designed for efficient partitioning of hypergraphs, particularly in the VLSI domain. The algorithm is based on the multilevel paradigm, which involves constructing a sequence of successively coarser hypergraphs, computing a bisection of the smallest hypergraph, and then refining this bisection to the original hypergraph. The algorithm's performance is evaluated in terms of the quality of the partitioning and run time on various VLSI circuits. The results show that hMEfS produces high-quality partitions that are 4% to 23% better than those of other state-of-the-art schemes, and it is significantly faster, often requiring 4 to 15 times less time. The algorithm scales well for large hypergraphs, with hypergraphs having over 100,000 vertices being bisected in a few minutes on modern workstations. The algorithm also outperforms other schemes consistently, especially for larger hypergraphs. The paper describes the framework of hMEfS, which includes a coarsening phase and an uncoarsening and refinement phase. In the coarsening phase, the algorithm reduces the size of the hypergraph by merging vertices or hyperedges. In the uncoarsening phase, the algorithm refines the partitioning by projecting the bisection to the next level finer hypergraph and using iterative refinement algorithms such as FM or HER to improve the quality of the partitioning. The algorithm uses two different refinement algorithms: FM, which moves vertices between partitions to improve the cut, and HER, which moves groups of vertices to remove entire hyperedges from the cut. The experimental results show that hMEfS produces better partitions than other algorithms, with improvements ranging from 4.1% to 21.4% in terms of hyperedge cut. The algorithm is also significantly faster, with run times that are 4 to 15 times less than those of other schemes. The algorithm is written in C and is available at http://www.cs.umn.edu/~karypis/metis. The paper concludes that the multilevel hypergraph partitioning algorithm is fast, robust, and effective for partitioning large hypergraphs.This paper presents a new multilevel hypergraph partitioning algorithm, hMEfS, which is designed for efficient partitioning of hypergraphs, particularly in the VLSI domain. The algorithm is based on the multilevel paradigm, which involves constructing a sequence of successively coarser hypergraphs, computing a bisection of the smallest hypergraph, and then refining this bisection to the original hypergraph. The algorithm's performance is evaluated in terms of the quality of the partitioning and run time on various VLSI circuits. The results show that hMEfS produces high-quality partitions that are 4% to 23% better than those of other state-of-the-art schemes, and it is significantly faster, often requiring 4 to 15 times less time. The algorithm scales well for large hypergraphs, with hypergraphs having over 100,000 vertices being bisected in a few minutes on modern workstations. The algorithm also outperforms other schemes consistently, especially for larger hypergraphs. The paper describes the framework of hMEfS, which includes a coarsening phase and an uncoarsening and refinement phase. In the coarsening phase, the algorithm reduces the size of the hypergraph by merging vertices or hyperedges. In the uncoarsening phase, the algorithm refines the partitioning by projecting the bisection to the next level finer hypergraph and using iterative refinement algorithms such as FM or HER to improve the quality of the partitioning. The algorithm uses two different refinement algorithms: FM, which moves vertices between partitions to improve the cut, and HER, which moves groups of vertices to remove entire hyperedges from the cut. The experimental results show that hMEfS produces better partitions than other algorithms, with improvements ranging from 4.1% to 21.4% in terms of hyperedge cut. The algorithm is also significantly faster, with run times that are 4 to 15 times less than those of other schemes. The algorithm is written in C and is available at http://www.cs.umn.edu/~karypis/metis. The paper concludes that the multilevel hypergraph partitioning algorithm is fast, robust, and effective for partitioning large hypergraphs.