The paper introduces Personalized Test-Time Scaling (PTTS), which treats test‑time computation allocation as a multidimensional optimization problem where users jointly specify accuracy, latency, and inference‑cost targets. Rather than optimizing a single resource dimension to push an accuracy‑cost or accuracy‑latency Pareto frontier, PTTS formulates the goal as discovering executable controllers that maximize the joint satisfaction rate of these user‑specific requirements. To amortize the cost of repeatedly searching for new controllers across diverse user profiles, the authors propose PersonTTS, an agentic policy‑discovery framework that reuses prior search experience. PersonTTS initializes each search with a controller whose requirements closely match the target user’s specifications and injects source‑distilled procedural guidance to steer the search, while still evaluating every candidate on the target profile to ensure fidelity. Experiments on the AIME and HMMT mathematical reasoning benchmarks demonstrate that PersonTTS substantially outperforms strong test‑time scaling baselines in joint requirement satisfaction on unseen user profiles and held‑out problems. Under a fixed candidate‑evaluation budget, reusing cross‑user experience not only improves the quality of discovered policies but also markedly reduces the discovery agent’s computational time and cost, confirming that amortized policy discovery can scale personalized test‑time scaling efficiently.