Agentic AI LLM Reasoning NLP MLOps
▸ML Engineer · research → production
Abdollahi

Mohammad

Portrait of Mohammad Abdollahi
▸ NOW

Bridging the gap between ML research and production.

Initialize contact
01

About

// profile

I'm a Machine Learning Engineer and researcher with 6+ years across Machine Learning, NLP, Large Language Models, Agentic AI and MLOps. I'm currently pursuing graduate studies at York University (Toronto, Canada) under the supervision of Prof. Hadi Hemmati (Google).

My research focuses on LLM reasoning and Agentic AI for software engineering — working on the theoretical foundations of AI to build intelligent, autonomous coding agents that solve complex software tasks.

I design scalable, end-to-end ML systems that close the gap between research and production. I don't just train models; I challenge them. My interest lives in the edge cases and the "impossible" problems where standard algorithms break down.

6+
Years in ML & AI
Multiple
Research papers in top-tier venues
02

Experience

// execution history
01
Machine Learning Engineer
Arteria AI — Toronto, CA
Visit
Sep 2025 — Jan 2026
  • Built an end-to-end RAG system using DeBERTa and vector search for legal documents.
  • Optimized Docker images, cutting size by 50% and accelerating build times by 70%.
  • Refactored and productionized ML products, ensuring CI/CD reliability.
02
Machine Learning Engineer
Digikala.com
Visit
Dec 2021 — Dec 2024
  • Redesigned the recommender system using ViT + BERT & HNSW, boosting conversion by 72%.
  • Designed an uplift-prediction tool that increased gross sales of daily deals by 39%.
  • Implemented LLM proofs-of-concept using the OpenAI API and open-source models.
03
Data Scientist
Snappfood
Visit
May 2021 — Dec 2021
  • Developed a sentiment-analysis system reaching 89% accuracy on noisy labels.
  • Improved food-categorization accuracy by 13% using fuzzy matching.
04
Software Engineer
Virgool
Visit
Jul 2019 — Aug 2020
  • Built a Persian text classifier (90% accuracy) and a spam-detection system (94% precision).
03

Stack

// installed modules

Languages

Python Go C / C++ Java JavaScript R

ML & AI

PyTorch TensorFlow HuggingFace OpenAI LangChain LangGraph Scikit-learn XGBoost Agentic AI RAG vLLM Ollama LMCache

Infrastructure & MLOps

Docker Kubernetes AWS GCP Kafka Airflow MLflow RabbitMQ Celery Databricks

Data & Backend

PostgreSQL MySQL MongoDB Redis Elasticsearch Qdrant GraphQL FastAPI Django
04

Research

// selected output
01
Demystifying Errors in LLM Reasoning Traces: An Empirical Study of Code Execution Simulation.
Mohammad Abdollahi et al.
02
Surveying the Benchmarking Landscape of Large Language Models in Code Intelligence.
Mohammad Abdollahi et al.
03
Deep-Bench: Deep Learning Benchmark Dataset for Code Generation.
Alireza Daghighfarsoodeh … Mohammad Abdollahi et al.
04
ComplexMatrixComb: Predicting Drug Combination IC50 Doses via Complex Numbers and Matrix Factorization.
Mohammad Abdollahi et al.
Google Scholar
05

Community

// service & outreach
Dec 2025 — Feb 2026
Program Committee & Reviewer
MSR 2026 — View committee
  • Rapid-response reviewer for the Mining Software Repositories conference (top-tier SE venue).
  • Delivered thorough, high-quality peer reviews under tight timelines.
Mar 2021 — Apr 2021
Scientific Consultant
Datadays 2021
  • Helped design a nationwide AI challenge on Learning to Rank and recommendation systems.
May 2018 — Aug 2020
Board Member
IUT Computer Engineering Student Scientific Association (CESSA) — GitHub
  • Organized seminars, scientific talks, and coding challenges to drive student engagement.
  • Ran workshops deepening students' understanding of computer engineering fields.
06 // let's build
Let's talk.