“model”

Model Training / Hyperparameter Optimization Engineer

Rigorous training loops: reproducibility, resource accounting, sane HPO, and checkpointing — for classical and deep models.

NoSQL/DynamoDB/Firestore Data Modeling

Designs access-pattern-first single-table/denormalized models — the opposite habit from relational design.

Model Validation / Experiment Designer (CV Strategy)

Chooses correct validation (grouped/time-series/nested CV), prevents over-opt, and reports uncertainty — the difference between local scores and prod results.

Model Evaluation, Fairness & Calibration Auditor

Beyond accuracy: calibration, subgroup performance, error types, and decision thresholds — what compliance/trust reviews actually need.

Model Monitoring / Drift Detection Designer

Production ML monitoring: data/concept drift, staleness, fairness over time, and retraining triggers — with alert thresholds tied to impact.

Model selection

Model Extraction and API Distillation Abuse Mitigation

防护商用/自研模型API遭高置信度查询蒸馏与克隆，设计速率管控、异常检测与水印追踪控制；适用于大模型推理服务运营方。

ER model in text (entities, attributes, types, nullability, PK/FK, cardinalities,

Ranking Model

3D model generation

Multi-Model Render Fleet Orchestration

在多家图像/图生视频/口型模型间做镜头级路由、级联与失败重试编排，兼顾质量、成本与交期。

Tension model: what creates risk/failure cost? Is failure legible and

Hit model: hurtbox/hitbox pipeline, active frame shape, multiple-hit rules,

Player-Model-Driven Adaptive World

基于实时玩家模型调节难度、节奏与生成式遭遇的多目标个性化系统，强调透明、可关闭、反暗黑模式。适用于服务型动作/射击/RPG。

AI models

Threat Modeling Workshop Facilitation

主持威胁建模工作坊,使用STRIDE/Attack Tree/PASTA方法识别系统威胁。

Data ownership migration: read model/API replacements for foreign readers,

Security Architecture / Threat Modeler (STRIDE)

Runs STRIDE on a feature/architecture before building: trust zones, abuse cases, and mitigations mapped to controls.

LLM Fine-Tuning / Model Customization Planner

Decides between prompting, few-shot, RAG, and fine-tuning; if fine-tuning, designs dataset, method (LoRA), and eval.

Multimodal Video Understanding Model Deployment for Film and TV

面向影视厂牌部署多模态视频理解模型，把每日素材自动转化为含镜头、人脸、台词、道具的结构化数据，打通场记、检索、质检与后期。