Modeling heat transfer in reacting-gas particle flows is important for understanding of biomass pyrolysis. In pyrolysis reactors, millions of ~1mm sized ground wood particles heat up quickly through convective heat transfer at the boundary layers at the particle surfaces. Modeling this correctly at the reactor-scale is a challenge due the irregular shapes of the particles, their low thermal conductivity, and vast number. At NREL we have taken a multi-scale approach involving: DNS of single particles with realistic geometry and a specially designed fixed bed heat-transfer experiment. Using a 1-D numerical model of the experiment, a minimization procedure was used to determine the heat transfer coefficients, particularly the particle Nusselt number¹s dependence on the Reynolds number. Using these results, the experiment was simulated using a 2-D transient model with the MFIX software. The focus of the talk is the multi-scale nature of the problem and the challenges encountered.