Cancer is one of the biggest medical challenges of our times, with a devastating impact across all social categories. Accordingly, considerable efforts are geared towards improving the standard of care. Among them, immunotherapy, which relies on the immune system's capacity to fight cancer cells, brings new hope for higher success rates and in some cases even a cure of this devastating disease. A particularly interesting approach called adoptive cell therapy, or ACT, is based on immune cells that have infiltrated tumors and have the potential to fight them. Unfortunately, these tumor-infiltrating lymphocytes (TILs) are very rare and their activity is often suppressed by components of the tumor microenvironment (TME). However, TILs can be extracted from tumor biopsies, reactivated and expanded in vitro before being reinjected into the patient. This immunotherapy strategy has shown impressive results in melanoma, where a majority of patients have seen a complete tumor remission. However, for most solid cancers (e.g. lung or colorectal cancer) that are in general less prone to mount immune responses, this approach is less effective. Other reasons for the currently limited success rate of ACT include the ineffectiveness to expand TILs in vitro, as well as the lack of functional assays to probe their potency after culture. Indeed, existing tumor models are based either on the non-physiological culture of cancer cell lines on plastic dishes or on the implantation of a patient's tumor tissue into an immune-deficient mouse, an assay that is highly complex, inefficient and expensive. Recent breakthroughs in 3D primary cell culture technology have provided paths forward in overcoming these challenges. In particular, the derivation of patient-derived miniaturized tumors, or tumoroids, has enabled the recapitulation of some key functional characteristics of real tumors. When co-cultured with fibroblasts or immune cells, tumoroids are potentially powerful models for immunotherapy applications. However, despite promising proof-of-concept demonstrations of modeling the TME with such heterotypic tumoroids, it has neither been possible to derive such complex in vitro tumor models with all key cell types isolated from the same patient's tumor, nor to maintain the numbers and phenotypes of TME-derived cells in longer-term cultures. Moreover, existing tumoroids are highly heterogeneous and unsuitable for the development of scalable and reproducible assays that are needed in diagnostic applications. This thesis focused on the bioengineering of novel in vitro tools to address some of the aforementioned obstacles in realizing the promise of TIL therapy. First, to generate heterotypic and autologous tumoroids, existing protocols were optimized to isolate and expand cancer cells, fibroblasts, and TILs from colorectal cancer (CRC) samples. A small biobank was established for each cell type as a resource for the development and application of patient-specific tumoroid assays. In the second part of this thesis, a heterotypic CRC tumoroid platform was developed for testing autologous TIL potency in combination with immunotherapeutic agents. CRC tumoroids were derived in high-throughput on a hydrogel substrate comprising an array of regularly spaced, round-bottom microcavities, a set-up that was found to highly amenable to quantitative experimentation. ... More in the full text ...