San Francisco, California, United States
My name is Nonye Okonkwo and I am a 3rd generation medical doctor and research scientist. Rather than pursue a career in private practice like most of my family, I decided to become a product manager and use technology to multiply my physical ability and bedside manners. I eat data for breakfast. I love reinforcement learning, designing schematics and optimizing agent decision making using peer-to-peer RL policy optimization. I am an unconventional future founder because I am homeless and pivoting from medicine to AI using only an iPhone 13 mini. I invested everything in retraining my medical-grade systems thinking to optimize for graceful trial and error learning and consistent chain of thought reasoning. I am a great fit for the AI application layer because I have a generalizable intelligence that will help enterprises scale and spread their research ambitions. I thrive in the gray area between immutability and statelessness. I intentionally seek out the social intersections where subjective and objective reality meet to understand the boundaries between AI vaporware and hardware utility. In my free time I build ambient computing interfaces that will one day redefine the way we love, think and work. I am the research scientist you have been looking for, Dwarkesh. I also happen to be a black woman who is well-versed in the racial, social and economic issues that will help us strengthen our codebase and scale the frontier of the AI application layer.
Advocating for UBI financial empowerment for residents displaced by AI in San Francisco and all formerly incarcerated Men and transgender Women of the 5 Keys Organization. Investigating post-model economics to design AI monetization strategies that leverage social media sentiment to drive AI adoption and subscription. Distilling market data, academic papers, and philosophical texts into engaging, lucid streams of consciousness optimized for intellectual clarity and action. Delivering core arguments and counter-intuitive questions through articles, letters and highly scannable, and rigorously cited research briefs, data deep-dives, and thesis-driven memos.
Coached clients on the importance of maintaining effective reward policy to combat inauthentic behaviors that arise during self-reported data collection. Examined patterns associated with high-quality self-analysis and designed productivity tools to identify and isolate actionable insights. Employed electronic health records data to develop robust insight pipelines, which played a key role in validating project cycles and tracking cognitive resource utilization. Designed proprietary scorecard tools to measure goal alignment and organize insights in behavior and productivity. Captured patterns linked to high-quality self-analysis and optimized a voice-based ambient user interface to promote effective self-optimization practices. Applied sentiment analysis techniques to trial and error language to optimize training loop schedules, context switching and mental bandwidth.
Researched the communication overhead of AI multimodality on usability scores and mobile product ergonomics. Designed adaptive feedback loops to enable discrete dialogues about performance and failure, delivering informal timely insights rather than waiting for formal reviews and 1-on-1’s. Reduced communication overhead associated with asynchronous personal knowledge management systems by implementing a robust labeling language and tracking temporal data devaluation. Led online discussions about dataset quality and passionately educated followers about the benefits of enrollment in human data research to improve AI output precision and utility.
Educated diverse audiences on the impact of data collection language on the curation of high-quality datasets, aiming to minimize bias from overfitting and oversampling. Leveraged LLMs like ChatGPT, Claude, LLaMa, Grok, Gemini, Deepseek and training platforms like OutlierAI to address cold-start issues in marketing content creation. Studied RLHF strategies to develop the personality of a conversational AI agent with a deep contextual understanding of user subjective and objective reality using natural language.
Improved model reasoning and domain-specific accuracy by applying natural language prompt engineering techniques that translated my subject matter expertise into broad instructions for desired LLM outputs. Performed pairwise comparisons of LLM model outputs to improve alignment with human intent. Gained hands-on experience with AI user interfaces and state-of-the-art models like ChatGPT and Claude. Actively participated in the core AI development lifecycle to garner a deep, practical understanding of how lagging-edge AI is built.
Utilized proprietary software applications to provide input and labels for classified projects. Played a pivotal role in contributing to the training of new tasks, collaborating closely with technical staff to ensure the successful design of efficient annotation tools. Engaged in diverse writing assignments to fine-tune generative AI models like Grok3.
Received extra training in pediatrics, geriatrics and bariatrics to establish a holistic healthcare worldview. Trained at the veterans hospital, nursing facility, geriatric home care, and pediatric clinics to improve intergenerational cultural competency. Designed an educational cartoon titled, the ADHD Brain Circuit Buddies to improve treatment outcomes for children with ADHD. Implemented a community engagement risk assessment intervention to improve resources allocation in a Spanish-speaking population.