Mark Torres

AI Engineer, Researcher, and Systems Builder

My name is Mark, an AI engineer, researcher, and systems builder. As a research engineer, I combine social sciences with AI tools, AI agents, and large-scale algorithms to study social media platforms and computationally redesign them to promote healthier discourse. I work at the intersection of AI research and methodology, developing and evaluating large language models as tools for computational social science. My work has been published in venues like Nature.

Outside academia, I also consult with teams across industries, helping startups, researchers, and organizations turn AI ideas into working systems.

Current: Lead AI Engineer @Northwestern (2023-present)

Building applied AI systems for social media research and computational social science.

Some highlights of my current work include:

  • Fine-tuning LLMs for custom content moderation research.
  • Building LLM-based simulations of social behavior.
  • Building platforms for reproducible AI + social science research.
  • Developing AI agents to enhance research methods and foster human+AI collaboration.
  • Exploring how generative AI can extend the methodological frontier of social science research.

Some past/ongoing work includes:

  • Designing and prototyping large-scale, AI-driven recommendation algorithms for experimental studies in political and social influence.
  • Built and maintained production data infrastructure ingesting and processing millions of social media records for academic research, including real-time data pipelines, ML inference services, and feed ranking algorithms that power multiple lab studies.
  • Led design and development of the ML and data pipelines for a large-scale, 2,000 user field study. Managed 10M daily records and hundreds of GBs of records. Developed and deployed recommendation and classification models. Managed orchestration using Prefect. Deployed on hybrid on-prem + AWS infra.

Current: AI consulting (2024-present)

Worked with multiple startups and small businesses across a variety of domains, helping teams go from “we want AI to do X” to a deployed agent workflow or live demo, whether X is RevOps, brand voice, or an entirely new product experience.

Sample projects:

  • AI agent workflow that automates the entire lead-to-meeting pipeline: real-time lead qualification (90%+ routing accuracy), automated CRM updates, and intelligent meeting scheduling. The system reduced response time from hours to under 10 minutes and converted 42% of routed leads to scheduled calls. (Demo link)
  • AI copywriting agent that generates customized marketing copy by analyzing a customer’s existing branding and messaging. The system maintains brand voice consistency while producing tailored content for campaigns, reducing content creation time while ensuring alignment with brand guidelines. (Demo link)
  • AI meditation coach agent that generates guided wellness exercises grounded in expert-defined best practices in relationship-building. (Demo link)
  • AI-powered Sports Broadcaster browser extension that adds a live, talking AI commentator on top of real games, reacting in real time to what fans are watching. (Demo link).
  • AI agent workflow that automatically monitors marketing campaign performance and sends natural-language alerts when performance drops are detected. Demonstrates workflow automation, integrating n8n, OpenAI GPT-4, and email/Slack for real-time business intelligence.(Demo link)

Previous: Data Scientist @SetSail (Series A startup) (2021-2023)

  • Built production NLP models for automated entity extraction and insight generation from sales communications (emails, call transcripts), enabling real-time analysis of thousands of conversations.
  • Architected agentic AI systems that integrated with Salesforce, HubSpot, Zoom, and Microsoft/Google ecosystems, automatically surfacing strategic insights from multi-channel sales data.
  • Developed ML-powered sentiment analysis models that analyzed prospect buying intent signals across email and call data, directly supporting revenue operations and pipeline forecasting.
  • Built predictive models for automated sales signal detection and deal scoring, using ML to identify high-probability revenue opportunities from patterns in sales activity data and customer engagement behaviors.
  • Developed behavioral scoring models that analyzed rep activity patterns (email frequency, meeting cadence, stakeholder engagement depth) to predict deal outcomes and identify deals at risk, directly informing sales strategy adjustments.

Previous: Data Scientist @Yale (2020-2021)

  • Developed a proprietary predictive algorithm for tracking the spread and incidence of COVID using network graph data (see Hunala). Served to tens of thousands of daily active users
  • Worked on large-scale deep learning models for identifying single-nucleotide polymorphisms in the human genome.

Published a few papers here and there

My Google Scholar Profile.

  • Brady, W.J., Doyle, M, Elnakouri, A., Finkel, E., Jackson, J.C., Kteily, N., Parker, V., Puryear, C., Spelman, T. , Teeny, J., & Torres, M. (In Principle Acceptance; registered report). Redesigning algorithms to intervene on social norm misperceptions during a national election. Nature.
  • Brady, W.J., McLoughlin, K.L., Torres, M.P. et al. Overperception of moral outrage in online social networks inflates beliefs about intergroup hostility. Nat Hum Behav 7, 917–927 (2023). https://doi.org/10.1038/s41562-023-01582-0
  • Iwamoto, S.K., Alexander, M., Torres, M. et al. Mindfulness Meditation Activates Altruism. Sci Rep 10, 6511 (2020). https://doi.org/10.1038/s41598-020-62652-1

Education

  • MS, Computer Science, @UT Austin
  • BS, Statistics, @Yale
  • Certificate, Mrs. Puff’s Boating School

Et cetera

I am a former digital nomad, now based out of Chicago.