Hello, I'm
Troy Arthur
Machine Learning Engineer & NLP/AI Researcher
Top-ranked globally at SemEval 2026 · 3+ years distributed systems · MS Computational Linguistics @ CU Boulder
01. About
I'm a Machine Learning Engineer at Wiland, Inc., where I've worked in this role since August 2026, with 3+ years of prior industry experience in distributed systems and data infrastructure. I'm also pursuing my MS in Computational Linguistics at the University of Colorado, Boulder — where I completed my BA in Psychology with minors in CS and Political Science.
My research focus is NLP and affective computing. In 2026 I competed in SemEval 2026, ranking competitively against 25+ international teams — including nationally funded research labs — and placing #1 globally on the Affective Valence prediction track, outperforming baseline by 24% with models trained entirely on a single local machine. My first-author paper on the work is published in the ACL Anthology.
My interests sit at the intersection of systems engineering and applied AI. I bring a resourceful, practical approach to research and engineering, with an emphasis on achieving strong results within real-world constraints.
Technologies
- Python
- PyTorch
- Hugging Face
- Spark / Scala
- C++
- TypeScript
- SQL
- HDFS
- Google Protobuf
- NumPy / Sklearn
- ChromaDB
- spaCy
- Linux Shell
- Git
02. Experience
Software Engineer
Wiland, Inc.Intern → Associate → Software Engineer
SparkScalaSQL
C++TypeScriptPython
HDFSProtobuf
Linux ShellGit
- Accelerated data pipelines through large-scale stack migration and distributed systems optimization.
- Optimized legacy C++ software, achieving significant improvements in system throughput and performance.
- Guided system architecture — produced ERDs and architecture diagrams to document and inform infrastructure design decisions.
- Security focus — hardened data infrastructure against unauthorized access and system vulnerabilities.
03. Education
MS in Computational Linguistics
University of Colorado, Boulder- GPA: 4.0 · CLASIC Scholarship (Sole recipient)
- Coursework: Natural Language Processing, Deep Learning, Advanced Algorithms, Morphology & Syntax
BA in Psychology, Minors in CS & Political Science
University of Colorado, Boulder- CS GPA: 4.0 · Latin Honors (thesis defense) · Undergraduate Research Grant
- Roles: Undergraduate Research Assistant (Stereotyping & Prejudice Lab); Course Assistant (Discrete Structures)
04. Projects
Affective State Prediction
Novel ensemble and hierarchical transformer model to predict user emotional states from text. Competed in SemEval 2026, ranking competitively globally and placing #1 on the Affective Valence prediction track, outperforming 25 international teams by 24% above baseline — with all models trained locally on a single machine. First-author paper published at SemEval-2026 (ACL Anthology).
PythonPyTorchHugging Face
NumPySklearn
CoRefRAG
Coreference-motivated RAG system introducing three novel discourse-aware document chunking strategies grounded in entity accessibility theory. Coreference chains inform chunk boundaries to reduce referential ambiguity in retrieved context. Evaluated over 412 Wikipedia articles from SQuAD 2.0 with custom chunking metrics and a full downstream QA pipeline.
PythonHugging FaceChromaDB
spaCyFastCorefClaude API
BERTScore
Research Data Aggregator
High-performance C++ data processing pipeline for a CU Psychology research team. Replaced brute-force Python parsing with a single-pass sliding-window algorithm, cutting processing time from an estimated 6 months to under 5 minutes. Custom BST for geographic CSV indexing.
C++R
05. Publications
-
NLPGroup8 at SemEval-2026 Task 2: Diverse Ensembles and Hierarchical Transformers for Emotional State Prediction
Proceedings of the 20th International Workshop on Semantic Evaluation (SemEval-2026), pages 379–386. Association for Computational Linguistics, 2026.
06. Contact
I'm always interested in connecting with others working in machine learning, NLP, and software engineering. For research discussions, collaboration, or professional connections, feel free to reach out.
Say Hello