> about
I'm an LLM engineer and researcher at Huawei Canada's Research Centre, where I work on coding agents: building scaffolds, designing evaluations, and developing benchmarks in one of the fastest-moving areas of applied AI. My broader work bridges NLP research with practical engineering to make LLM-driven systems more capable, controllable, and reliable.
I completed my M.A.Sc. in Electrical and Computer Engineering at Queen's University in September 2025, where my thesis (CoDial) introduced an interpretable framework for aligning LLM behaviour with explicit dialogue flows. The work was published at ACL 2026. I have additional publications at ICML 2026 and EMNLP 2024.
Alongside research, I contribute actively to open-source LLM tooling: I authored the FeatBench adapter and the trae-agent integration for Harbor (Terminal-Bench's official harness), and have four merged PRs to NVIDIA's NeMo Guardrails covering parser fixes, prompt rendering, and developer tooling.
experience
Associate LLM Engineer · Huawei Canada
Works on coding agents at Huawei Canada's Research Centre. Concrete contributions:
- Improved a coding-agent scaffold, boosting performance by >15% on SWE-Bench Verified.
- Analysed coding-agent outputs on LiveCodeBench through statistics and plots, surfacing issues both in model reasoning (reported back to the data team to inform the next SFT iteration) and in the evaluation harness itself.
- Lifted gold-patch resolution on FeatBench from 37.31% (±1.88) to 99.81% (±0.10) by diagnosing harness and environment defects in the original benchmark. Released FeatBench-Verified v1.3 to the community and ported parity fixes upstream to the original FeatBench repo.
Big Data Engineer · Bale Messenger
Built and maintained production data infrastructure:
- Developed and maintained robust big data pipelines for continuous data ingestion and processing, enhancing data reliability and real-time accessibility.
- Implemented scalable solutions with Apache Spark, Hive, and Kafka, significantly improving data-warehousing efficiency and scalability.
ML Engineer & Backend Developer · Sharif Data Analytics Lab & National Elites Foundation of Iran
Two threads on a voice-command recognition product:
- Trained a speaker-verification system leveraging the state-of-the-art ECAPA-TDNN model, achieving over 97% accuracy on a custom 8,400-pair dataset.
- Designed and developed a web API using a microservices architecture with Flask and Laravel, streamlining voice-command processing and authentication.
Scientific Team Member · SamCode 5 Programming Contest
Designed algorithmic problems and comprehensive test cases for a programming contest targeting high-achieving junior-high students.
Freelance Developer · Ponisha
Built Telegram-based applications in Python using Telegram's TDLib for a range of clients.
education
M.A.Sc., Electrical and Computer Engineering · Queen's University
- Thesis: CoDial: Interpretable Task-Oriented Dialogue Systems Through Dialogue Flow Alignment (supervisor: Prof. Xiaodan Zhu). Published at ACL 2026.
- Research areas: NLP, dialogue systems, NLP in the legal domain, machine learning.
B.Sc., Computer Engineering · Amirkabir University of Technology
- Ranked 1st among 156 students.
- B.Sc. project: Generating Handwritten Persian Characters with Generative Models (supervisor: Prof. Ahmad Nickabadi).
- Selected coursework (all 20/20 unless noted): Computational Intelligence, Artificial Intelligence, Data Structures and Algorithms, Information Retrieval, Programming Languages, Signals and Systems, Advanced Programming, Web Programming, Multicore Programming (19.8/20), Multimedia Systems (19.5/20), Data Mining (19.5/20), Operating Systems (19.4/20).
Diploma, Mathematics and Physics · Allame Helli 3 High School (NODET)
- NODET = National Organization for Development of Exceptional Talents (Iran's network of selective schools).
skills
Honest gaps (not claimed as expertise): distributed training (FSDP, DeepSpeed, Megatron), RLHF/DPO/GRPO post-training, and inference servers (vLLM, SGLang) — surface familiarity only.
teaching
Queen's University, Department of Electrical and Computer Engineering
- Teaching Assistant, Machine Learning & Deep Learning (Winter 2025, Prof. X. Zhu)
- Teaching Assistant, Fundamentals of Information Structure (Fall 2024, Prof. T. Dean)
- Teaching Assistant, Object-Oriented Programming (Winter 2024, Prof. M. Greenspan)
Amirkabir University of Technology, Computer Engineering
- Teaching Assistant, Multicore Programming (Spring 2022 & 2023, Prof. M. Momtazpour)
- Teaching Assistant, Computational Intelligence (Spring 2022, Prof. M. Ebadzadeh)
- Head Teaching Assistant, Signals and Systems (Fall 2021, Prof. M. Rasti)
- Teaching Assistant, Signals and Systems (Spring 2021, Dr. A. Aghaeeyan)
- Teaching Assistant, Data Structures and Algorithms (Fall 2021, Prof. E. Nazerfard)
- Teaching Assistant, Linear Algebra (Spring 2021, Prof. M. H. Chehreghani)
certifications
- AWS Technical Essentials — Amazon Web Services, Spring 2025
- Generative Adversarial Networks Specialization — DeepLearning.AI, Spring 2022
- Game Theory — Stanford University & The University of British Columbia, Summer 2021
- Deep Learning Specialization — DeepLearning.AI, Summer-Fall 2020
- Machine Learning — Stanford University, Spring 2020
- Reinforcement Learning (audited) — University College London (David Silver), Summer 2020
honors & awards
- Top 0.3% among 130,000+ applicants — Nationwide University Entrance Exam (B.Sc. in Mathematics and Engineering), Iran (2018)
- 3rd place, Iran Zamin Open Cup (IZOCup) Programming Contest — Held by Salam High School, Tehran (2014)
- Qualified, NODET entrance exam — National Organization for Development of Exceptional Talents, Iran (2011)
languages
- Persian: Native
- English: Proficient. TOEFL (2022): 109 / 120 (R: 29, L: 30, S: 21, W: 29).
References available on request.