ShawnL

- Multi-Agent Systems & Workflow Orchestration – self-evolving agent ecosystems, dynamic workflow generation (WorkFlowGenerator, AgentManager), and reinforcement-driven coordination strategies. - Optimization Algorithms for Agents – experimentation with MIPRO, MASS, TextGrad, and AFlow to optimize prompt distributions, emergent reasoning strategies, and agent communication topologies. - LLM-Centric Reasoning & Benchmarking – fine-tuning and evaluating LLM-driven agents on GAIA, HotPotQA, and MBPP, targeting measurable gains in reasoning depth and execution reliability. - Data-Driven Financial AI – automated data ingestion (akshare), multi-scale feature engineering, and integration of technical indicators (MA, RSI, MACD, Bollinger Bands) for algorithmic trading pipelines. - Hybrid Human-in-the-Loop Systems – approval-gated interceptors, adaptive feedback loops, and semi-automated validation layers for safe decision augmentation. - Adaptive & Generative AI – workflow morphogenesis, representation learning, and algorithmic prompt evolution for scalable open-source research.