EXPERIENCE

Machine Learning Software Engineer (PhD Researcher)

BC Cancer Research Institute | September 2021 – Present

• Train 100M+ parameter Transformers on TB-scale, long-context (>10k tokens) multimodal data (PyTorch).
• Design and scale self-supervised representation learning models for high-dimensional, terabyte-scale data.
• Engineer C++ inference engines (libtorch, OpenMP), cutting runtime from days to 2.7h and RAM usage by 3x.
• Quantize models and train on multiple GPUs to improve throughput by an order of magnitude (CUDA, SLURM).
• Architect and maintain the full software development lifecycle for open-source C++ tools (10,000+ downloads).
• Ship production-ready pybind11 and libtorch bindings with automated CI/CD release pipelines (Docker, Azure).
• Spearhead core ML R&D, yielding a provisional U.S. patent, 4 first-author papers, and grant funding.

Machine Learning Researcher (Part-Time)

University of Isfahan | September 2016 – September 2021

• Engineered an end-to-end ML pipeline for user modeling, predicting user traits from social media text data.
• Trained deep reinforcement learning agents in NLP environments utilizing NLTK and fuzzy logic.
• Instructed ML, NLP, Python, and Django courses for the ACM Student Chapter (~20 students per cohort).
• Built an automated grading platform (Python) using sockets, with C++ and Java interfaces for project evaluation.

Software Engineer (Contract)

University of Isfahan | September 2018 – September 2019

• Built a Django + PostgreSQL social platform for 100,000+ users with an automated credential verification engine.
• Administered the production Linux (Ubuntu), Nginx, and uWSGI stack, ensuring high availability for daily traffic.

SKILLS

ML/AI: Recommender Systems • User Modeling • Search & Ranking • Representation Learning • NLP • RL • LLMs
Languages/Frameworks: Python • C++ • PyTorch • libtorch • Django • PostgreSQL
Systems/Inference: CUDA • OpenMP • pybind11 • TorchScript • Model Quantization • Memory Optimization • SLURM
Developer Tools: Cursor / GitHub Copilot • Docker • Git • CMake • CI/CD • Linux

EDUCATION

PhD in Bioinformatics

University of British Columbia | September 2021 – November 2026

• Focus: Multimodal Machine Learning, Inference Optimization, and Sequence Modelling (PI: Dr. Inanc Birol)

M.Sc. in Computer Engineering

University of Isfahan | September 2019 – June 2021

• Focus: Deep Learning, Language Modelling, Reinforcement Learning, and Natural Language Processing

B.Sc. in Computer Engineering

University of Isfahan | September 2015 – June 2019

• Competitions: RoboCup 2D Soccer Simulation, ACM-ICPC West Asia Regionals, Computer Engineering Olympiad

PROJECTS

Myrid: Self-Supervised Deep Learning for Large-Scale Signal Data Processing

Provisional U.S. patent filed in 2026 | PoC presented at RECOMB 2026

• Built an end-to-end data processing pipeline and representation learning system for TB-scale temporal data.
• Trained a self-supervised model, using quantization for memory- and compute-efficient inference.

Fuzzy Word Sense Induction and Disambiguation

Paper: 10.1109/tfuzz.2021.3133905

• Designed a context-aware semantic classifier using fuzzy clustering on word embeddings (BERT, word2vec).
• Validated the architecture against standard natural language processing benchmarks (SemEval 2010 and 2013).

AIEdit: Deep Learning for Alignment-Free CER Improvement

GitHub: BirolLab/AIEdit | Paper: 10.1371/journal.pcbi.1014245

• Engineered a production-grade inference engine, compiling PyTorch to C++ via TorchScript and pybind11.
• Achieved a 58% error reduction, outperforming the next-best baseline while strictly optimizing latency.

ntStat: Toolkit for Statistical Analysis of Sequence Data

GitHub: BirolLab/ntStat | Paper: 10.1371/journal.pcbi.1014158

• Implemented n-gram counting via Bloom filters in C++ (pybind11), with 99.5% accuracy at reduced memory usage.
• Applied differential evolution and BFGS optimization to accurately model token count histograms.

ntHash2: High-Throughput Rolling Hash Algorithm for Pattern Matching

GitHub: BirolLab/ntHash | Paper: 10.1093/bioinformatics/btac564

• Implemented an optimized algorithm in C++ and shipped high-performance Python bindings via SWIG.
• Accelerates hashing throughput by 3.8× over conventional algorithms for pattern matching (spaced seeds).

SELECTED TALKS

• TEDx University of Isfahan: Not So Artificial Intelligence
• Vancouver Bioinformatics User Group (VanBUG): Modelling k-mer profiles with evolutionary algorithms

VOLUNTEER EXPERIENCE

Student Mentor & Conference Adjudicator | UBC | 2024 – 2025

• Mentored 5 students through UBC’s Undergraduate Research Opportunities program.
• Adjudicated undergraduate research posters at UBC’s Multidisciplinary Undergraduate Research Conference.

Volunteer Organizer | Vancouver Bioinformatics User Group, VanBUG | 2023 – 2024

• Coordinated community engagement for seminars by local and international researchers.