Software Engineer and Machine Learning Researcher
I build reliable software systems and study how large language models reason about programs. My current interests include LLM reasoning, program semantics, efficient generative models, and AI-enabled software systems.
- 💼 Software Engineer at Argoman, leading the development of HR technology products
- 🎓 M.Sc. student in Data Mining at Shahid Beheshti University
- 🔬 Researching reliable and verifiable LLM reasoning about program transformations
- 🧮 B.Sc. in Mathematics and Applications from Amirkabir University of Technology
Google Scholar · LinkedIn · Medium
A training-free method for accelerating diffusion-model inference by adaptively merging similar tokens and caching token-pair information across denoising steps.
My work focused on reducing redundant self-attention computation while preserving image-generation quality.
A research benchmark for evaluating whether large language models can reason reliably about semantic relationships between programs.
Rather than evaluating only binary equivalence predictions, EquiBench studies whether models can:
- Identify semantics-preserving and semantics-altering transformations
- Explain why program behavior is preserved or changed
- Produce consistent reasoning across related examples
- Generate explanations that can be checked systematically
Status: Research in progress.
An educational deep-learning framework created to explain how systems such as PyTorch work internally.
Mini-Torch implements concepts including tensors, automatic differentiation, computation graphs, neural-network modules, and optimization using a compact, readable architecture.
My master’s research investigates the reliability of large language models in reasoning about semantic relationships between programs.
The work develops a structured evaluation framework for determining whether LLMs can move beyond simple equivalence classification and provide reasoning that is consistent, structured, and verifiable.
This research is conducted in collaboration with Dr. Khashayar Etemadi.
I lead a software-engineering squad developing HR-related products. Our work translates business requirements into maintainable systems using Domain-Driven Design principles.
I am also exploring how AI agents can improve product workflows, automate repetitive operations, and support more intuitive user experiences.
I worked at Yektanet, one of Iran’s leading advertising-technology companies, contributing to high-throughput distributed systems.
My work included OpenRTB-based SSP integrations, CTR prediction services, video-advertising systems, internal tools, and production monitoring using technologies such as Python, Kafka, PostgreSQL, Redis, gRPC, Sentry, and Grafana.
- M.Sc. in Data Mining, Shahid Beheshti University
- B.Sc. in Mathematics and Applications, Amirkabir University of Technology
- Lead Teaching Assistant for Data Mining, Computational Data Mining, Deep Learning, Linear Algebra, and User Interface Design
An implementation-focused explanation of decision-tree construction and visualization.
I am interested in research and engineering collaborations related to:
- Reliable LLM reasoning
- Program analysis and code intelligence
- Efficient machine-learning systems
- AI-assisted software products