I am building a cinematic blast sequence and need solid data-driven support rather than eye-balling particle emitters. The idea is to run an AI/ML simulation that predicts how metal, glass, concrete, wood and plastic fragments fly away from the detonation point, then turn those numbers into clear data-visualisation charts I can hand directly to the FX team. Scope • Direction, speed and distance must be predicted separately for each material class, following the colour code I already use on set (red = metal, blue = glass, grey = concrete, orange = wood, green = plastic). • The charts have to highlight the mean and standard deviation for every variable so the artists immediately see the “typical” trajectory as well as the natural spread. Other metrics such as percentiles or min/max can stay in the background; we will only display them later if the director asks. • A short Python/TensorFlow (or PyTorch) pipeline is preferred so we can regenerate results when the blast parameters change. Feel free to leverage NumPy, Pandas and Plotly or Matplotlib for plotting—whatever keeps the workflow light and editable. • Final delivery should include the source code, a reproducible notebook, the generated CSV/JSON of raw predictions and the rendered charts (PNG or interactive HTML). Acceptance A run of at least 500 simulated blasts must complete in under 10 minutes on a standard RTX-class GPU, and the resulting charts need to show physically plausible decay of velocity with distance for each material. If the mean paths line up with our reference footage within ±15 %, the job is done. Designed and developed by.. JAVED AKHTER ANSARI