Aspiring AI/ML Engineer
Tyler
Katz
Building AI agents, RAG systems, and fine-tuned LLMs.
About
I'm Tyler, a graduate student at Syracuse University working toward a career in AI/ML engineering. I got here through data: my undergrad in Applied Data Analytics at the iSchool taught me to find the signal, and somewhere between my first scikit-learn pipeline and my first LLM agent, I realized I cared most about building the systems that put models to work.
I just wrapped up a data science internship at Comcast on the security infrastructure analytics and orchestration team, where I automated a quarterly network inventory reconciliation process that used to take tens to hundreds of hours by hand. The pipeline uses an LLM to compare inventory records across multiple source systems, generate structured reports, and reconcile discrepancies between databases, running through FastAPI endpoints I built and MCP integrations I contributed to that let agents handle the process automatically. That work now takes one to two minutes instead of hundreds of hours.
I'm currently a full-time graduate student, still chasing that same problem: turning slow, manual processes into systems people can just trust.
On campus, I lead engineering for United AI, Syracuse's AI/ML student organization, where I help other students go from curious to shipping.
I also spent a year as a data analytics research assistant, building the tables, charts, and maps behind a professor's book on the relationship between crime and politics in Brazil. I was a major contributor to that work, and a few of the maps I built ended up presented in front of the UN at a conference.
Education
2026
B.S. Applied Data Analytics
Syracuse University — iSchool
In Progress
M.S. Applied Human Centered Artificial Intelligence
Syracuse University — iSchool
Skills
- Languages
- PythonJavaScriptRSQLHTMLCSS
- ML & Data Science
- scikit-learnpandasNumPymatplotlibPySpark
- AI / LLM Engineering
- LangGraphRAGAgentsLLM Fine-Tuning (LoRA/QLoRA)Prompt Engineering
- Engineering & Cloud
- FastAPIDockerAWS EKSGitHub Actions CI/CDStreamlitShiny
Experience
Data Science Intern
June – Aug 2026
Comcast — SIAO, Cybersecurity Research Engineering
- Automated quarterly network device inventory reconciliation by building an LLM agent and MCP integration, cutting the process from hundreds of hours to 5 minutes and reducing token costs 30–40%.
- Engineered a FastAPI pipeline across 14 device platforms to auto-generate and upload CMDB templates, improving data consistency for ~5,000 network devices.
- Applied the agent to generate structured discrepancy reports across 14 platforms, cutting total reporting time from hundreds of hours to 2–3 hours.
Engineering Lead
Aug – Dec 2025
United AI — Syracuse University
- Led 10 student AI/ML teams, meeting weekly to unblock issues and guide projects to production.
- Served as technical lead, advising on Python and model training/deployment, and reviewing designs for code quality.
- Managed a 10-project GitHub org (repos, CI/CD) and lectured 200+ members on Git, Claude Code, and new AI tools.
Data Analytics Research Assistant
Jan – Dec 2025
Syracuse University
- Engineer clean, integrated datasets from raw census files, geospatial shapefiles, and web-scraped sources using R, streamlining data preparation workflows and improving analysis efficiency for the project.
- Design and implement data visualizations, geospatial maps, and statistical tables that were directly incorporated into multiple chapters of Machine [Gun] Politics, enhancing the clarity and persuasiveness of research findings.
- Translate complex, multi-source datasets into actionable insights through advanced data wrangling, exploratory analysis, and visualization techniques, driving progress toward book completion and supporting evidence-based conclusions.
Projects
Agentic RAG System
A LangGraph research assistant over 2,299 chunks of job postings and arXiv abstracts that decides when to retrieve, reformulates weak queries, and cites its sources — not another 'chat with your PDF' clone.
outcome: Query-reformulation agent beat a simple RAG pipeline on answer faithfulness (2.80 vs. 2.75/3) at a 90% retrieval hit-rate, proven with a custom eval harness.
- AI / LLM
- LangGraph
- RAG
- FastAPI
- Chroma
- Docker
Fine-Tuned LLM Microservice
A small open LLM fine-tuned with QLoRA to extract structured job-posting data, then merged, quantized, containerized, and shipped to a live AWS endpoint behind a full CI/CD pipeline — not a notebook demo.
outcome: Fine-tuning lifted tech-stack extraction from 0.08 to 0.68 F1 over the zero-shot baseline, after diagnosing and fixing a training-data regression that had initially made the model worse.
- AI / LLM
- QLoRA
- FastAPI
- Docker
- AWS EC2
- GitHub Actions
Fake News Detection
An NLP classifier that scores a news article's credibility from its title and body text, stacking a logistic regression and a random forest into a meta-classifier and shipping the result as an interactive Streamlit app.
outcome: 95.51% accuracy from a stacked meta-classifier — beating both individual base models on every metric on a 20% held-out test split of 72,134 articles.
- Classical ML / DS
- scikit-learn
- NLTK
- TF-IDF
- Streamlit
- GridSearchCV
Sales Forecasting Predictions
An XGBoost pipeline that forecasts daily store-level sales for 1,115 Rossmann retail locations from 2.5 years of history, served through a Shiny dashboard for comparing actual vs. predicted sales by store and time window.
outcome: Feature engineering and hyperparameter tuning together cut RMSE by ~54% (1433.87 → 663.47) and MAE by ~59% (1081.46 → 438.57) versus the baseline model.
- Classical ML / DS
- XGBoost
- R
- Shiny
- ggplot2
- tidyverse
Customer Segmentation
A PySpark pipeline that clusters 2,240 retail customers into behavioral personas with K-Means, then builds three recommendation engines — association-rule mining, a hybrid recommender, and rule-based deal-matching — on top of the segments.
outcome: Chose K = 3 over the elbow method's suggested K = 6 by weighing three validation methods together — the PCA projection was the tiebreaker that a metric alone would have missed.
- Classical ML / DS
- PySpark
- MLlib
- K-Means
- FP-Growth
- PCA
Resume
The one-page version
- M.S. Applied Human Centered AI @ Syracuse — in progress
- Data Science Intern @ Comcast
- LangGraph agents, RAG, and LoRA/QLoRA fine-tuning in production settings
- Engineering Lead of United AI, Syracuse's AI/ML student org
- Data Analytics Research Assistant — data visualizations and geospatial analysis for a published book