Matthew T. Dunn
I’m Matthew Dunn, serving as Head of AI at Resurface Therapeutics on an interim basis. The company is developing lipid nanoparticle-based medicines; as an investor, I’m helping build its AI and scientific computing foundations while it gets established, including protein language models for LNP design and optimization and AI-assisted research platforms.
Previously, I co-founded and led Daedaline, served as Research Director and Member of the Technical Staff at Cohere, and built and led LivePerson’s applied AI team.
Earlier at NYU, I worked on question answering and language modeling, ported gated recurrent unit (GRU) code from Theano to PyTorch, and used one of the first DGX-1 clusters deployed there—well before today’s LLM boom.
Experience
In an interim capacity, help establish the company’s AI and scientific computing capabilities, developing protein language models for LNP design and optimization and AI-assisted platforms for literature analysis and scientific diligence.
Raised $2M in seed financing, built the founding team, and developed a production AI platform for real-asset due diligence, including retrieval, distributed inference, and agentic query routing.
Set annotation and forward-deployed engineering strategy and prototyped model adaptation methods for production language model applications.
Led a 20-person applied research team and company-wide deployment of deep learning across conversational AI products.
Built and led an applied NLP research team delivering production language models for LivePerson Conversational Cloud.
Helped build the applied machine learning team and developed NLP methods for large conversational datasets.
Worked on the SearchQA dataset and open-domain question answering in the CILVR group.
Developed simulation, forecasting, and probabilistic bidding models for U.S. electricity capacity markets.
Built a laboratory information management system for high-throughput genetic sequencing diagnostics.
Industry Reviewer
Research
- Actionable Conversational Quality Indicators for Improving Task-Oriented Dialog Systems (2022)
- Domain-specific Distillation Models for Conversational Commerce (2022)
- SearchQA: A New QA Dataset Augmented with Context from a Search Engine (2017)
- Early Predictability of Asylum Court Decision (2017)
- Optimizing Targeted Sequencing of Somatic Mutations (2016)
Patents
- Domain Adaptation of AI NLP Encoders with Knowledge Distillation · US-20220318502-A1
- Automatic Bot Creation Based on Scripts · US 10,853,568 B2
- Intent Driven Contact Center · US 11,050,884 B2
- Systems and Methods for Intent Analysis · US 10,902,199 B2
Education
- M.S. Data Science, NYU Courant Institute · CILVR Lab / ML2 Group
- B.A. Environmental Economics, University of Redlands · Minor in English Literature
Volunteering
Advisor to NYU Center for Data Science capstone groups (2019–present).