Optimal dietary protein and lipid proportions during early fish ontogeny are shaped by protein deposition, non-protein energy supply, digestive maturation, and the physiological consequences of nutrient excess. This paper develops a controlled-rearing factorial design and a reproducible computational experiment to examine joint dietary protein–lipid effects on specific growth rate, feed conversion ratio, protein efficiency, and larval survival. A 3 × 3 design spans 40%, 45%, and 50% protein with 8%, 12%, and 16% lipid, yielding 45 independent synthetic tanks (five per diet), each followed at days 7, 14, 21, and 28. Tank-held-out linear, ridge, random forest, and histogram gradient-boosting models were compared for specific-growth-rate estimation. In the artificial experiment, the 50% protein/16% lipid treatment yielded the largest mean day-28 specific growth rate (3.965% day−1) and survival (94.08%), while the 50%/12% combination produced the lowest feed conversion ratio (1.340). Random forest achieved the smallest held-out RMSE (0.170), but these differences reflect programmed response functions rather than actual larval physiology. A new conceptual figure links diet composition, larval digestive and energetic processes, water-quality controls, and measured outcomes. The study provides a transparent computational template for designing species-specific protein–lipid feeding trials, with explicit attention to tank-level replication, uncertainty, competing nutritional endpoints, and biological validation.