Exploring heavy flavour physics at a hadron collider offers numerous opportunities, though it comes with significant demands on the trigger and reconstruction due to the high multiplicity environment. The recent upgrade of the LHCb detector introduces a fully software-based trigger to handle higher signal rates. Its real-time reconstruction algorithms must consider speed, performance and bandwidth, while at the same time accommodating a diverse physics program. Beyond enhancing the regular program, the upgraded setup opens new avenues, from low-mass di-electrons from hypothetical dark photons, to directly reconstructing charged beauty hadrons. In this presentation, the speaker will give an overview of the real-time reconstruction efforts, with a focus on their work: from electron and charged beauty hadron reconstruction to machine learning infrastructure.