The next decade will see the arrival of a new generation of instruments with a large increase of data flux at the 2020 horizon, and beyond. Among these, LSST will provide a unique information in the visible band and it will be a major actor to probe transient phenomena: supernovae, gamma ray bursts, microlensing, mergers, asteroids, and more. LSST is expected to produce each night millions of alerts of transient phenomena candidates, and this daunting volume of alerts is one of the major differences with respect to previous experiments. To fully harness the power of these instruments, new methods must be developed to deal with large data volumes and to coordinate resources for follow-up of promising candidates.
In this talk I will present Fink, an IN2P3 initiative being developed to face the big data challenges of the LSST alert stream. Fink exhibits traditional astronomy broker features and focuses on a large set of science cases with a large part dedicated to supernova and multi-messenger astronomy. It also goes beyond these features by providing real-time transient classification which is continuously improved by using state-of-the-art Deep learning and Adaptive Learning advances. These evolving added values will enable more accurate scientific output from LSST photometric data and maximize the scientific return on both LSST and collaborating experiments.