I am a Computer Science graduate and AI/ML researcher. I have worked on research projects at
Macquarie University and Amirkabir University of Technology, and I have backend engineering
experience from industry. My next step is a research-focused graduate path, and I am eager to stay
engaged with international research activity.
Research Interests
Natural Language Processing, Large Language Models, Vision-Language Models,
Agentic AI, Information Retrieval, Uncertainty Quantification.
Languages
English - TOEFL iBT 93 / 120 Professional Working Proficiency ·
Persian - Native
Current Research
Research Intern, University of Electronic Science and Technology of China Sep 2026 - Present · China | Remote Supervised by Prof. Chaoning Zhang
When Neighbors Mislead: A Retrieval Framework for Misinformation
Detection Saleh Shakoor, Behnam Yousefimehr, Mehdi Ghatee Manuscript in preparation · 2026
When Neighbors Mislead: A Retrieval Framework for Misinformation
Detection Saleh Shakoor, Behnam Yousefimehr, Mehdi Ghatee Manuscript in preparation · 2026
Research Experience
Research Intern, University of Electronic Science and Technology of
China
Sep 2026 - Present
China | Remote · Supervised by Prof. Chaoning Zhang
Details
Building foundational knowledge of vision-language models (VLMs) as part of
research onboarding.
Research Intern, Macquarie University
Apr 2025 - Apr 2026
Centre for Applied AI, Sydney, Australia | Remote · Supervised by Prof. Amin
Beheshti
Details
Built a retrieval-based framework for misinformation detection and
credibility assessment, requiring no model fine-tuning.
Reached 84.9% F1 on a frozen Mistral-7B, beating a
77.9% zero-shot baseline and a 74.8% semantic-RAG
baseline.
Addressed semantic-RAG failure in deceptive content detection with a
veracity-oriented retrieval key.
Engineered an indexing scheme compact enough to cover 8,000+ docs in under
25 KB.
Transferred the framework across models, raising Qwen3-8B F1 from 70.4% to 80.0%
via signals from a stronger model.
I wanted to build real international research
experience. The Centre for Applied Artificial Intelligence stood out for its
active, fast-moving environment, so I reached out to Prof. Amin Beheshti and
asked for an internship. I had the option to join an ongoing graduate project or
design my own, and I chose to design mine around Trustworthy and Explainable AI,
an area I care deeply about. I noticed that semantic retrieval based methods
often failed at misinformation detection and content credibility assessment, so
I set out to design a more reliable way to retrieve similar cases for this task.
It was my first real experience in an international research setting, and it
laid the groundwork for my future career.
Undergraduate Research Assistant, Amirkabir University
Feb 2025 - Aug 2025
Innovation and Data Science Lab | On-Site · Supervised by Dr. Mohammad
Akbari
Details
Replaced the pipeline with a unified LLM slot-filler that
robustly parses complex expressions into structured JSON.
Diagnosed colloquial and context-dependent Persian phrasing as the key
limitation of a legacy rule-based slot extraction pipeline.
Enabled reliable intent extraction and structured information generation
through few-shot prompting in a single model call.
Built an end-to-end time and task management agent integrated
with Google Calendar API for automated scheduling.
When I decided to start my journey into academic AI
research, joining the Innovation and Data Science Lab under the supervision of
Dr. Mohammad Akbari turned out to be one of the best opportunities I could have
asked for. The lab always had active projects, and students worked closely
together across every stage of the work, so undergraduates like me could learn
directly from graduate researchers. When I joined, the lab's main project was a
Persian language intelligent assistant built around three modules: media
recommendation, time and task management, and information retrieval. I was
assigned to the time and task management module.
Bachelor's Thesis
Sep 2024 - May 2025
Optimization of Vehicle Routing Problems with Variable Constraints ·
Supervised by Prof. Mostafa Shamsi
Details
Benchmarked Local Search & Metaheuristic pairings across 5 VRP variants
using OR-Tools.
Showed GLS and Tabu Search recover optimal solutions from suboptimal
starts.
Found hard constraints significantly degrade Metaheuristic performance.
Before moving into AI, I worked as a backend developer,
and one of my most meaningful roles was at a company building a navigation and
maps product. That experience sparked my curiosity about vehicle routing
problems, so I chose it as my thesis topic and worked on it under the
supervision of Prof. Mostafa Shamsi. It gave me a deep appreciation for local
search and metaheuristic methods, and it's where my research instincts really
started to take shape.
Backend Developer, Neshan
Maps(Developed and maintained backend services for 30M+ clients at 1,100+ req/sec)
Oct 2023 - Apr 2025
Details
Developed and maintained backend services for 30M+ clients at
1,100+ req/sec peak flawlessly.
Redesigned caching strategy, cutting p99 latency from
2s to 18ms, and reducing database read load.
Centralized session management systems with JWT,
LDAP & Spring Boot to enable independent
scaling.
Implemented async messaging using RabbitMQ to decouple services
and improve resilience.
Set up monitoring with Prometheus and Grafana to
ensure service uptime.
Designed database schemas optimized for high-throughput
workloads across distributed services.
As a computer science student, I was eager to get real
industry experience beyond what academia could offer, since a lot of practical
software skills just aren't something you pick up in a classroom. I also had a
strong drive to go deep on software engineering, because I knew implementation
skills matter as much in research as they do in industry. That's how I joined
Neshan Maps through a backend development bootcamp. Neshan is a navigation and
maps product that guides millions of Iranians to their destinations every day.
Interestingly, global players like Google Maps and Apple Maps have a much
smaller footprint in Iran, mainly because Neshan is built around local needs and
feels more natural for people to use.
Built and maintained automated data processing pipelines by integrating
third-party APIs.
Designed microservice architecture for fault isolation and independent
scalability.
Early in my studies, a group of students from my
department started GoldScan, a fintech startup. I had no real industry
experience at that point, and I wanted to put myself in a fresh, fast-moving
environment, one that could become a real starting point for my own growth.
