Brain monitoring is pivotal in medical diagnostics and treatment, offering real-time insights into brain function that empower healthcare professionals in the effective management of various brain disorders. However, traditional monitoring methods like microdialysis-based or electrochemical sensor-based systems suffer from limitations such as low temporal resolutions, insufficient selectivity, and poor stability. Recent advancements in optical sensing, microfabrication, signal processing, and machine learning present innovative solutions to surmount these constraints and enhance monitoring capabilities. Therefore, this thesis aims to develop an optical fibre sensing system for implantable and multiplexed brain monitoring. In its initial phase, a reflection optical fibre-based sensing system was engineered to enable the continuous monitoring of four brain biomarkers—pH, oxygenation, glucose, and temperature—simultaneously. Leveraging optical sensing techniques and novel film fabrication methods, highly sensitive and selective biosensors were developed and seamlessly integrated onto the tip of a reflection fibre for four distinct brain biomarkers monitoring. To further improve the sensing performance, in the subsequent stage, an optical fibre bundle-based sensing system was proposed, aiming to incorporate two additional biomarkers, Na+ and Ca2+, while reducing the sensor size for minimally invasive measurements. Polymer-encapsulated fluorescent receptors were adopted for the sensor fabrication to enhance the sensing sensitivity and reversibility. Characterisation test results affirmed the sensors' high sensitivity, specificity, reversibility, and environmental robustness in measuring all six biomarkers. Mathematical modelling algorithms, including machine learning and deep learning models, were employed for spectrum processing, environmental compensation, and precise biomarker concentration readings. The proposed sensing system successfully identified three commonly encountered brain complications accurately using ex vivo brain models. In the end, a fully automated sensing system was prototyped by integrating the optical components and algorithms, enabling continuous online brain biomarker monitoring. The proposed multiplexed brain monitoring system underwent successful validation using real human cerebrospinal fluid samples, showcasing its potential for clinical application.