AI / ML Engineer | Computational Intelligence | Research-Driven Builder

I’m a senior at Salisbury University graduating in May 2026, focused on building intelligent systems that do more than fit data. My work lives at the intersection of machine learning, optimization, symbolic regression, computer vision, and LLM-supported decision systems.

I’m especially interested in problems that require reasoning under uncertainty, strong validation, and real-world deployment. I've already heavily explored spaces from financial-model discovery pipelines to medical triage agents and vision systems for applied detection/counting.

• Building AI/ML systems that combine modeling, search, validation, and practical software engineering
• Exploring symbolic regression + grammatical evolution for financial time-series structure discovery
• Developing LLM-supported healthcare triage workflows with probabilistic dialogue refinement
• Creating computer vision pipelines that turn raw model outputs into more reliable downstream predictions

1st Place — Rutgers Health Hack 2025 | Continued development contract | Sole AI/ML engineer

Built an AI-supported OB/GYN triage system designed to interpret patient language, refine predictions through guided dialogue, and support subspecialty routing. I developed the core AI/ML logic, including custom NLP-style interpretation and probabilistic conversation guidance, and later helped evolve the system toward an LLM-supported deployment pitch.

Core themes: NLP, decision logic, uncertainty refinement, healthcare workflow support, deployment-minded prototyping

Undergraduate research | ~10k LOC private repository

Developed a Python pipeline using NumPy and Numba to automatically generate, optimize, validate, and group predictive models for financial data. The system explored model structure discovery at scale and used Shapley-value-based ensemble methods to combine strong candidates into a more coherent forecasting framework.

Core themes: genetic programming, symbolic regression, ensemble systems, Monte Carlo validation, time-series research

Data Science Capstone | In development

Current capstone project focused on combining symbolic regression and grammatical evolution to study how structured search spaces can improve efficiency in time-series model generation.

Core themes: search-space design, grammatical evolution, symbolic regression, computational experimentation

Computer vision project for applied detection/counting

Trained an Ultralytics YOLO model on complex visual data for object detection and counting, then built downstream meta-models that used YOLO emissions as structured features. Validation showed substantial reduction in both prediction error and error variance.

Core themes: computer vision, object detection, post-model feature engineering, statistical validation

Designed a multi-stage neural-network experimentation pipeline that rotates across feature/time partitions, trains multiple models, and aggregates emissions into higher-level meta-model outputs.

Core themes: ensemble architecture, time-series modeling, feature segmentation, neural-network experimentation

Languages
Python, C, C++, Bash, Java, JavaScript, HTML, R, SQL

Libraries / Frameworks
NumPy, Numba, Pandas, TensorFlow, Keras, PyTorch, scikit-learn, Ultralytics

Focus areas
Machine Learning, Genetic Programming, Symbolic Regression, Model Validation, Computer Vision, Monte Carlo Methods, Optimization, Data Structuring

Tools
Git, Linux, WSL, VS Code, CUDA

• President, AI Club at Salisbury University
• Computer Science tutor and lab assistant across 100–400 level coursework
• NCAA All-American in discus
• 2x NCAA championship qualifier, school-record holder, and multi-time conference champion

I care a lot about disciplined iteration, building from first principles, and turning messy real-world problems into systems that can actually be tested and improved.

This profile includes a mix of coursework repositories, experiments, and smaller public projects. Much of my strongest current AI/ML work is in private research or team repositories, so if you’re interested in a project mentioned here, feel free to reach out.

• Email: 7logankelsch5@gmail.com
• LinkedIn: linkedin.com/in/logan-kelsch

I’m currently looking for opportunities in AI/ML engineering, applied machine learning, and software engineering where I can contribute as a builder, researcher, and fast learner.