Sean KellyKelly Enterprises

Work

What I've built, and what it taught me

Platform architecture and applied AI in production today, preceded by consulting, a couple of ventures, and electrophysiology research. The through-line is that the hard part is almost never the model — it's the structure of what you're reasoning over.

Experience

  1. Chief Technology Officer · Incite.ag

    Apr 2024 — present

    Sterling, IL

    A hands-on technical leadership role — roughly 70% of my time is still in production code, architecture, prototyping, and applied R&D. I own the architecture and implementation of every technical layer of the SaaS platform, from application development through data infrastructure, cloud and hybrid systems, partner integrations, production operations, and security.

    • Architect core domain models and data pipelines against USDA FD-CIC technical guidelines and evolving 45Z requirements — translating regulatory guidance and scientific research into auditable agricultural reporting data.
    • Research and build agent-driven document-intelligence pipelines applying OCR, NER, and structured extraction to a corpus exceeding 10,000 pages annually, mapping unstructured and semi-structured PDFs and scanned records into proprietary schemas.
    • Scale production ingestion and analytics across 10M+ agricultural acres, thousands of farm operations, and dozens of biofuel organizations each representing more than 100M gallons of annual fuel production.
    • Architect secure hybrid integration frameworks connecting farm management systems, on-premises plant infrastructure over S2S VPNs, and partner platforms through APIs — with identity-aware workflows, traceable transformations, and resilient data pipelines.
    • Own cybersecurity architecture, implementation, and testing across the platform: identity, data protection, audit-ability, secure development practices, and production risk management.
    • Manage and coach several direct-report engineers through architecture reviews, code reviews, and technical problem-solving, while translating rapid regulatory and market change into executable engineering strategy.
    PythonAgent pipelinesOCR / NERData infrastructureHybrid cloudSecurity architecture

  2. Technology Consultant, Next Generation AI · EY

    Sep 2022 — May 2024

    Boston, MA

    Built internal and external generative AI strategy and the systems underneath it, across a mix of client delivery and firm-wide platform work.

    • Implemented vector databases and knowledge graphs for large-scale semantic search, scraping and structuring large text and numerical datasets out of enterprise documents.
    • Fine-tuned open and closed source LLMs on transformed text corpora in Azure, and built recursive LLM-driven data analytics using code interpreters and APIs.
    • Led large-scale internal AI initiatives in the emerging technology engine, including work with EY's QA leadership to transform contract management for one of the firm's largest business functions.
    • Helped design and steer development of EY's generative AI platform (EY.ai), iterating on the forward-looking roadmap with leadership and technical alignment across development teams.
    • Built machine learning time series demand forecasting models for large manufacturing clients.
    • Presented strategic considerations and research findings to clients and senior EY leadership.
    LLM fine-tuningVector databasesKnowledge graphsAzureTime series

  3. Founder · Massachusetts Move Insights

    May 2021 — Mar 2023

    South Bend, IN

    A venture built around automating the compilation and delivery of client lead information out of a private database — and around a computer vision problem nobody else was bothering to solve.

    • Automated end-to-end compilation and delivery of client lead information from a private database.
    • Classified properties as furnished or unfurnished using TensorFlow image classification CNNs on AWS infrastructure.
    TensorFlowCNNsAWSData pipelines

  4. Co-Founder, Business Development · Cogitare Engineering

    Dec 2021 — Sep 2022

    South Bend, IN

    Technical, problem, and solution validation for information extraction pipelines — plus market sizing and commercialization strategy.

    • Validated technical approach and market fit for extraction pipelines built on natural language processing, ontologies, knowledge graphs, and semantic web technologies.
    NLPOntologiesKnowledge graphsSemantic web

  5. Senior Risk Assessment Analyst · IDEA Center, University of Notre Dame

    Sep 2021 — Sep 2022

    South Bend, IN

    Assessed the commercial potential of technology coming out of Notre Dame research labs, and managed the analysts doing the same.

    • Managed 3 technical market analysts, meeting weekly to guide their individual risk assessment reports.
    • Determined market potential and opportunity for technology discovered by ND researchers — including graph neural networks for drug discovery, drone-assisted augmented reality, and chemiresistive sensors.
    Technology assessmentMarket analysisGNNs

  6. Researcher · College of the Holy Cross, Dept. of Neuroscience

    Jun 2019 — Jun 2021

    Worcester, MA

    Electrophysiology and psychophysics — applying machine learning to EEG, and running perception experiments on human subjects. This is where the habit of treating biology as a signals problem got started.

    • Applied multivariate pattern analysis (MVPA) to EEG data to classify event-related potentials, and to identify the nature of the information carried by conventional electrodes versus tri-polar concentric ring electrodes.
    • Implemented perception illusion paradigms and analyzed psychophysical data in MATLAB.
    EEGMVPAMATLABSignal processingPsychophysics

  7. Founder · Kelly Enterprises LLC

    2023 — present

    Boston, MA

    The entity behind independent consulting work, side projects, and this site.

    ConsultingML systems

Education

  • M.S. in Engineering, Science & Technology Entrepreneurship (ESTEEM) — AI at Scale · University of Notre Dame

    May 2022

    Notre Dame, IN

    A graduate program pairing technical depth with venture formation. My thesis work went into deep neural network interpretability — a novel method for assessing multi-way feature explainability against a new ground truth measure, which became a first-author AAAI 2023 paper.

  • B.A. in Multi-Disciplinary Neuroscience · College of the Holy Cross

    May 2021

    Worcester, MA

    Coursework distributed across computer science, mathematics, biology, and physics, all feeding the neuroscience department’s electrophysiology research. Graduated with a 3.45 GPA. The longer story of how that degree came together is in the writing section.

The longer version of that story is in Why I went looking for a major that didn't exist.

Publications

  1. 2023

    Feature Instability Search for Multi-Way Explainability

    Kelly, S., & Jiang, M.

    AAAI 2023 Workshop on Representation Learning for Responsible Human-Centric AI (R2HCAI)paper

    Proposes feature instability, the distance an input feature must move to flip a model’s output, as a proxy for feature importance, alongside DFEST, an approach for quantifying multi-way feature interactions. Together they define a synthetic ground truth for explainability, so explanations are evaluated against a known answer rather than against other explanation methods.

    PDFCode

  2. 2021

    Multivariate pattern analysis of EEG from conventional and tri-polar concentric ring electrodes

    Kelly, S., et al.

    Journal of Vision — Vision Sciences Society (VSS) 2021abstract

    Applied MVPA to EEG data to classify event-related potentials and to characterize the information carried by conventional electrodes versus tri-polar concentric ring electrodes.

    DOI

  3. 2020

    Perception illusion paradigms and psychophysical analysis

    Kelly, S., et al.

    Journal of Vision — Vision Sciences Society (VSS) 2020abstract

    Implemented perception illusion paradigms and analyzed the resulting psychophysical data in MATLAB.

    DOI

Projects

8 entries

Tools & certifications

Languages
Python (Selenium, Qiskit)JavaScript / Node.jsC / C++JavaMATLABPrologWolfram / Mathematica
AI / ML
LLMsAgentsModel evaluationInformation extractionNEROCRTensorFlow / PyTorchCNNsGPU-accelerated inferenceMVPAGNNsConstraint satisfaction
Systems & infrastructure
DockerContainerized model servingDistributed data pipelinesCloud-native & hybrid architectureProduction observabilityPrivate npm & PyPI registriesAPI & integration architecture
Cloud & security
AzureAWSIdentity & access managementTenant isolationGovernanceSecure developmentSecurity testing
Knowledge representation
PandasPySparkOntologiesKnowledge graphs / Neo4jEmbeddings
Certifications
Neo4j Certified ProfessionalNeo4j Graph Data ScienceMicrosoft Azure Data FundamentalsMicrosoft Power BI Data Analyst

Elsewhere