member of technical staff/ai research · aws applied ai
currently working on amazon bio discovery ↗
a research platform that consolidates a project's code, documents, and audio and makes the whole corpus queryable through retrieval-grounded conversation. I designed and built it end to end on aws (bedrock, retrieval, dynamodb, lambda), with a full test suite. in beta, in active evaluation with two teams as of march 2026.
betaan approach to making agentic systems reliable enough for scientific use. control flow is fixed in code, as directed acyclic graphs (dags) and logic flow graphs (lfgs), rather than chosen by the model, which must ground each step in retrieved evidence before acting.
in experimentationan on-device facial-analysis pipeline (raspberry pi), built as the visual front-end for experiments on how prosody and facial expression carry meaning together.
activebefore birth, a baby hears its mother's voice through the womb, which filters speech down to its melody and rhythm. this project recreates that muffled signal and asks whether a small, brain-inspired network trained only on it can still tell one language from another, and speech from non-speech, the way newborns can.
in design/experimentationa brain-inspired approach to the forgetting problem in retrieval-augmented generation: a two-system memory modeled on how cortex and cerebellum divide labor in mammals. under internal evaluation at amazon.
paper in prepa position paper, in preparation, on developmental cognition as a target for general intelligence.
position paperI'm an applied scientist working on deterministic ai systems for life sciences.
I design and build the systems, adapt the models inside them, and develop the experimentation and evaluation that test how well they hold up. at aws applied ai I built resonance, a research platform now in beta with two teams; earlier I led the research behind amazon rufus (now used by 250 million people, with over $10b in incremental sales) and behind alexa's speech and language systems. my background is in philosophy, neuroscience (columbia; harvard's martinos center), and computational linguistics (university of washington), with a focus on the development of language, intelligence, and perception. my own research concerns prosody, memory, and how children learn.
education & training
past industry
stack
book and paper summaries.
short audio summaries of the books and papers I'm reading. recent: yang & chen on neuromorphic intelligence, wegner on the illusion of conscious will, arendt on the banality of evil; earlier, ricoeur on time and memory, marcus on soutine, and adorno's minima moralia in full.
dratsi-pod · subscribe via rss
industry case studies available on request.