CV

Yehor Mishchyriak

ymishchyriak@wesleyan.edu | LinkedIn | ymishchyriak.com

Education

Wesleyan University, Middletown, CT

BA in Computer Science and Mathematics (double major) | May 2026

Cumulative GPA 3.63/4.00 | Mathematics GPA 3.55/4.00 | Computer Science GPA 3.67/4.00

Relevant Coursework

• Mathematics — Differential, integral, and multivariable calculus, statistics, probability, discrete mathematics, linear algebra, applied linear algebra, abstract algebra, real analysis, differential equations, combinatorics
• Computer Science — Foundations of computer science I & II, algorithms and complexity, computer networks, artificial intelligence, audio-visual machine learning, SQL and databases, HPC, quantitative textual analysis, automata theory, program analysis

Research Experience

Bonham Lab, Tufts University School of Medicine (Boston, MA)

Student Research Intern (remote, part-time) | November 2025–Present

• Developing and evaluating graph neural networks for semi-supervised protein function classification on large-scale protein co-occurrence graphs derived from human gut metagenomic profiles.

Summer Research Intern (on-site, full-time) | June–July 2025

• Built an end-to-end ML data engineering pipeline for protein functional annotation (data mining, cleaning, feature engineering, persistence).
• Encoded categorical variables as multi-hot vectors with dimensionality reduction. Embedded amino acid sequences using the ESM2 protein language model and free text using the OpenAI SDK with caching.
• Implemented Zarr-based persistence and packaged the pipeline into a documented Python library, M2F (microbiome-to-function).
• Processed ~1k microbiome metagenomic profiles output by HUMAnN, yielding ~900k protein entries with numerically encoded UniProt annotations, compressed for ML.

Main tools — NumPy, Pandas, PyTorch, PyG, Hugging Face, OpenAI, Zarr, Tufts HPCC

Thayer Lab, Wesleyan University (Middletown, CT)

Research Assistant (hybrid, part-time) | September 2024–Present

• Rebuilt and optimized residue interaction network (RIN) construction and random-walk sampling pipelines developed during my summer fellowship, adding modular design, logging, error handling, shared-memory parallelization, and careful memory management. Added RIN visualization and MD-trajectory animation with granular control over residues, interaction thresholds and types, and visual settings.
• Developed an embedding pipeline (PyTorch + my NumPy-based framework PureML) using DeepWalk-style random walks with interaction-weighted transition probabilities and Skip-Gram training to learn context-rich amino acid embeddings, including PCA projection and visualization. Packaged the pipeline into the SAWNERGY Python library with tests and CI/CD.
• Applying SAWNERGY to analysis of the p53 tumor suppressor protein and its isoforms to discover effects of mutations and ligand binding on the protein.

Research in Science Fellow (on-site, full-time) | June–July 2024

• Studied biophysics, molecular biology, Unix-based systems, and molecular dynamics simulations.
• Developed the initial pipeline to construct RINs from MD trajectories and sample random walks, enabling amino acid co-occurrence analysis and comparison between systems (e.g., mutant vs. wild-type) to characterize mutation- and effector-induced changes in allosteric signaling.

Main tools — NumPy, Matplotlib, Zarr, PyTorch, PyG, PureML, Wesleyan HPCC

Publications and Preprints

• Mishchyriak, Y. PureML: A Transparent NumPy-Only Deep Learning Framework for Teaching and Prototyping. Journal of Open Source Software, 2025. DOI 10.21105/joss.09631.
• Mishchyriak, Y., Stetson, S., and Thayer, K. SAWNERGY: A Python Framework for Dynamic Residue-Interaction Networks and Walk-Based Embeddings from Molecular Dynamics Simulations. Manuscript in revision for resubmission to the Journal of Open Source Software following updated requirements. DOI 10.5281/zenodo.18512892.
• Caballero Mancía, F., Stetson, S., Mishchyriak, Y., and Thayer, K. Isoform Context Shapes the Structural and DNA-Binding Impact of the p53 Y220C Mutation. Journal of the American Chemical Society, under review.

Preprints are available at ymishchyriak.com/papers.

Presentations

• Wesleyan Summer Symposium (2024) — Presented the initial pipeline for constructing RINs from MD trajectories and sampling random walks.
• Annual Biophysics Retreat (Wesleyan Biophysics Program) (2024, 2025) — Presented work on residue interaction networks in the context of allosteric signaling (2024). Presented SAWNERGY and its applications to downstream machine learning and p53 isoform analysis (2025).
• MB&B209 Guest Lecture (2025) — Delivered a lecture on allosteric regulation in proteins and Thayer Lab’s research in this context.
• ACS Northeast Regional Discussion (2025) — Presented SAWNERGY and its applications to downstream machine learning and p53 isoform analysis.

Projects

PureML (NumPy-based deep learning library)

Personal Project | Creator and Maintainer | May 2025–Present

• Built a lightweight deep learning framework in NumPy with a graph-based reverse-mode autograd engine, Tensor API, modular layers (Affine/Linear, Embedding, BatchNorm, Dropout), activation functions, loss functions, and batched training utilities.
• Informed the framework’s design and implementation by systematically working through the first 15 chapters of Deep Learning by Goodfellow, Bengio, and Courville, translating key theoretical concepts into reproducible code.
• Implemented optimizers (SGD, AdaGrad, RMSProp, Adam) with learning-rate schedulers, deterministic seeding utilities, and full checkpointing of model, optimizer, and scheduler state via compressed Zarr-based archives for reproducible experiments.
• Achieved ~98% test accuracy on MNIST with a small fully connected network implemented entirely in PureML, matching classic PyTorch baselines and validating correctness and numerical stability of the framework.
• Packaged and released PureML as an open-source Python library with logging, tests, and documentation. Actively extending the framework and using it as a backend in other research codebases such as SAWNERGY.

Main tools — NumPy, Zarr

eQoScan (WesHack, Wesleyan University Hackathon)

Personal Project | Developer | November 2024

• Built a mobile/web app for Wesleyan’s reusable food container system with real-time tracking (holder details, pick-up and drop-off times, locations) and automated return reminders.
• Integrated dynamic QR codes for seamless dining hall transactions.
• Developed in 24 hours and selected as a winner among 70+ U.S. university participants.

Main tools — Flask, Firebase, JS/HTML/CSS/Bootstrap, Swift (iOS)

Additional Work Experience

Hazel QAC and Department of Math and CS, Wesleyan University (Middletown, CT)

Applied Vectors and Matrices Teaching Assistant (on-site, part-time) | January–May 2025

• Led weekly help sessions for the class of 30 students. Assisted during in-class laboratory sessions.

Differential Calculus Teaching Assistant (on-site, part-time) | January–May 2024

• Led weekly help sessions for the class of 25 students. Graded homework assignments.

Instructional Media Services, Wesleyan University (Middletown, CT)

On-Campus Events Assistant Manager (on-site, part-time) | December 2023–May 2024

• Set up and managed AV equipment for on-campus events across multiple venues attended by dozens.
• Managed an average of 10–12 events per week.
• Composed weekly schedules for 18 student workers.
• Provided support and guidance for student workers.
• Enhanced my ability to respond quickly to challenges under time constraints.

Classroom Support Technician (on-site, part-time) | September 2022–May 2024

• Provided faculty and students with live technical support.
• Performed maintenance on electronic equipment, such as computers, multimedia projectors, screens, network gear, microphones, speakers, and cameras.

Skills and Knowledge

• Programming Languages — Python, C, OCaml
• Python Ecosystem — NumPy, Pandas, Matplotlib, PyTorch, PyG, Zarr, Numba, Cython, Flask
• HPC — Unix systems, Slurm, Bash, parallelization (threading/multiprocessing, IPC), memory management
• Tools — Git, SQL
• Languages — English (near-native), Ukrainian/Russian (native), German (elementary)