Independent AI Researcher (Causal Mechanism Design & RLHF Tuning)
TextRLHF
I led independent AI research projects focused on designing and evaluating causal mechanisms in neural architectures. My work involved iterative AI training, including aligning generative models and evaluating their performance via benchmarks. I implemented RLHF-based feedback to refine LLM outputs and research prototypes.
• Engineered a 114M parameter Differential Attention head evaluated on Diagnosis Arena.
• Performed evaluation and red-teaming of LLMs for safety and causal refusal ("Causal Bomb" detection).
• Utilized Hugging Face Transformers and RLHF methodologies for training/feedback cycles.
• Documented alignment strategies and efficiency benchmarking of SOTA research models.
I led independent AI research projects focused on designing and evaluating causal mechanisms in neural architectures. My work involved iterative AI training, including aligning generative models and evaluating their performance via benchmarks. I implemented RLHF-based feedback to refine LLM outputs and research prototypes. • Engineered a 114M parameter Differential Attention head evaluated on Diagnosis Arena. • Performed evaluation and red-teaming of LLMs for safety and causal refusal ("Causal Bomb" detection). • Utilized Hugging Face Transformers and RLHF methodologies for training/feedback cycles. • Documented alignment strategies and efficiency benchmarking of SOTA research models.
2023 - Present