Stephan Hähne

• ● Prototyped agentic AI workflows using LangChain and LangGraph.

• ● Built LLM-powered systems integrating GPT-4, Claude, and open-source models with custom business logic and workflow orchestration.

• ● Conducted independent research and experimentation in latent chain-of-thought models and their theoretical underpinnings, inspired by the latest ICML/NeurIPS work on mechanistic reasoning.

• ● Contributing to open-source benchmarking, code replication, and synthesis of insights between machine reasoning and cognitive neuroscience.

• ● Designed and deployed a MILP-based planning system for capacity allocation and workload scheduling across German and Chinese manufacturing facilities, reducing production costs by 15–20% and lowering lead times.

• ● Partnered directly with the CPO and engineering teams to frame executive-level operational challenges (capacity, orders, cost constraints) into tractable mathematical models, successfully transitioning prototypes from Python to Kotlin production systems.

• ● Implemented a simulation framework to evaluate system behaviour under increased demand, enabling scenario analysis and capacity stress-testing, the same analytical pattern underlying wait-time.

• ● Developed SQL-based analysis of optimization outputs and tradeoffs to support management decision-making, with results visualized in operational dashboards.

• ● Implemented SLAM-based computer vision algorithms (OpenCV, PyTorch) for sensor measurement, demonstrating an ability to contribute across engineering domains adjacent to core OR work.

• ● Analyzed large-scale revenue and impression data to inform bid-price optimization across connected ad-auction platforms, representing an early form of real-time marketplace pricing.

• ● Built the technical infrastructure to support automated pricing APIs, revenue data pipelines, and ad configuration systems (Python) connected to external platforms reducing system error susceptibility by 20%.

• ● Systematized legacy implicit optimization logic for pricing, revenue modeling, and ad configurations into transparent, scalable frameworks, empowering management with independent decision-making capabilities.

• ● Translated complex data patterns into clear trade-offs and strategic recommendations for the CEO, helping shape product roadmap priorities and technical scaling decisions.