This work is about eliminating lemmas from constructive proofs. Proofs are represented using intuitionistic sequent calculus and lemmas are cut inferences. So in the end we are actually eliminating cuts from intuitionistic sequent calculus proofs. The way we do this is via a method based on resolution. Why do we want to eliminate lemmas? Because it is important for proof analysis. Different proofs of the same theorem might give us different insights on the theorem itself. We use intuitionistic sequent calculus because we are interested in constructive proofs, which, in principle, give us more information than a non-constructive proof. A constructive proof can actually provide a concrete algorithm for solving a problem. We use a method based on resolution inspired by CERES, which is cut-elimination by resolution for classical logic. CERES can perform better than reductive cut-elimination, complexity-wise speaking, on some types of proofs. Also, the resulting proof after eliminating the cuts via CERES can be different than the ones reductive cut-elimination will yield, which means more information for mathematical proof analysis. Our main goal is to develop a CERES-like method for intuitionistic logic such that it has the same advantages as CERES for classical logic. The CERES method, as is, for classical logic cannot be straightforwardly applied to intuitionistic calculi. The main difficulty comes from the fact that intuitionistic calculi are often defined by some structural restriction on the sequent. These restrictions are bound to be violated when performing some of CERES' operations and because of this the resulting proof tends to be classical. We need to modify the method such that it accommodates these restrictions and yields intuitionistic cut-free proofs. This change can be done in many ways along the steps of CERES, and several methods were developed. Our first solution is based on the application of extra inferences to avoid the violations. This turned out to work only for a subclass of intuitionistic logic. It involves using a new resolution calculus and redefining the elements of CERES. The second solution works for the same subclass, but with the original elements of CERES and a simple post-processing of the final proof. The third approach required an extensive study of rule permutations in the sequent calculus for intuitionistic logic LJ. It is based on the observation that if the input proof is of a specific shape, the final proof can be transformed into an intuitionistic proof by an operation we define. The rule permutations are necessary to transform the input proof into this shape, and interestingly enough, some unsoundness is allowed. This unsoundness comes from the permutation of strong quantifier inferences. Usually, these are not allowed for they might generate eigenvariable violations, but in our case, we can live with the violations and guarantee that they will not occur on the final proof. The fourth and fifth methods are not methods per se. The former defines a new way to assemble the final proof in CERES and the latter is a resolution refinement that approximates CERES to reductive cut-elimination. We conjecture that, by combining these operations we obtain a cut-elimination method by resolution for intuitionistic logic (without strong quantifiers in the theorem to be proven). This thesis summarizes the main results in developing a method of cut-elimination by resolution for intuitionistic logic. Besides the methods already mentioned for sub-classes of this logic, we also discuss the relation between CERES' operations and intuitionistic logic, and pinpoint exactly where the problem lies. We finish the thesis with a conjecture that such a method is possible. The proof of which we leave as future work.