Database Systems
Retrieval and query machinery inside production engines — graph-based RAG in GaussDB, subgraph matching, and graph query languages.
GaussDBNeo4jSQLC++
Huawei HKRC · Hong Kong
Mohammad Matin Najafi
Database & ML Research
Ph.D. in Computer Science and Senior Software Engineer building research systems across database engines, machine learning, and AI agents — and shipping the ones that work into production.
About
Research that reaches production
I am a Senior Software Engineer on the database team at Huawei Hong Kong Research Center, working across database systems, machine learning, and AI agents. I hold a Ph.D. in Computer Science from The University of Hong Kong, with publications at VLDB, ICDE, and CIKM.
My work sits between research and engineering: prototype in Python, move hot paths into C++, and carry the result into a real system. Recent projects include graph-based retrieval inside GaussDB, token-efficient memory for coding agents, and autonomous memory under edge–cloud constraints.
Before Huawei I was a postdoctoral fellow at HKU and a visiting scholar at UIUC and the University of Kiel. Away from work I play competitive chess and research quantitative markets.
- Database Systems
- Machine Learning
- AI Agents
- Graph Learning
- Quantitative Finance
Research
Four threads I keep pulling on
Machine Learning
Graph neural networks and their limits: subgraph counting benchmarks, one-shot pruning for sparsification, and representation learning on graphs.
PyTorchGNNsBenchmarking
AI Agents
Memory and context for agents that operate at scale — hierarchical code memory, context gating, and autonomous memory across edge and cloud.
LLM SystemsRetrievalEdge–Cloud
Quantitative Markets
An independent research interest: time-series foundation models for forecasting, systematic execution, and options-implied densities.
ChronosTime seriesOptions
Experience
Where the research ships
-
Senior Software Engineer
Huawei Hong Kong Research Center · Database team
Research in database systems and AI agents
- Designed and productionized a graph-based RAG system for associative, multi-hop retrieval; filed a patent and integrated the system into GaussDB.
- Received Huawei’s FutureStars Medal of Honor, a peer-nominated award recognizing high-potential employees.
- Contributed to ContextSniper, a hierarchical code-memory and context-gating system that cut token use by ~39–52% on matched SWE-bench Lite tasks.
- Developing autonomous memory for agents under resource-constrained edge and cloud environments.
- Lead a four-person engineering team; review designs and code; maintain shared testing and research practices.
PythonC++RustGitCI/CD -
Postdoctoral Fellow
The University of Hong Kong · with Prof. Reynold C.K. Cheng
- Research on graph neural networks and graph databases.
- Supervised two Ph.D. students.
PythonPyTorchC++Neo4jGraphQL -
Visiting Scholar
University of Illinois Urbana-Champaign · with Prof. Kevin C.-C. Chang
- Graph mining and database research contributing to ICDE and VLDB publications.
- Visited database groups at Harvard, MIT, and UCLA.
PythonPyTorchC++ -
Visiting Scholar
University of Kiel · with Prof. Matthias Renz
- Efficient algorithms for discovering motifs on labelled and dynamic graphs.
- Conducted under the Germany / Hong Kong Joint Research Scheme.
PythonC++Algorithms -
Software Engineer Intern, Backend
Cafe Bazaar · Android marketplace with 40M users
- Built Mini-inlines, a platform answering informational search queries directly inside the marketplace.
- Built Crawlport, a crawler collecting and processing live sports data.
- Containerized search services and monitored them with Prometheus and Grafana.
PythonDjangoDockerKubernetesJenkinsRedisMongoDBAWS -
Earlier research & engineering
HKU · IPM · Andisheh Computer
- HKU (2019) — Summer research intern on motif discovery in heterogeneous information networks.
- IPM (2018) — Research intern on code-layout optimization with Dr. Pejman Lotfi-Kamran and Dr. Ali Ansari.
- Andisheh Computer (2017) — Built a business-intelligence Android app for Iran’s National Power Authority.
Publications
Selected papers
2026
2025
2025
Empowering Embarrassingly Simple One-Shot Pruning for Graph Sparsification
Graph sparsification & pruning · manuscript
2025
MuSha: Subgraph Matching by Multilevel Sharing
IEEE ICDE 2025 · pp. 2548–2561
2024
Large Subgraph Matching: A Comprehensive and Efficient Approach for Heterogeneous Graphs
IEEE ICDE 2024 · pp. 2972–2985
2024
ZeroEA: A Zero-Training Entity Alignment Framework via Pre-Trained Language Model
PVLDB 17(7) · pp. 1765–1774
2023
2020
M-Cypher: A GQL Framework Supporting Motifs
ACM CIKM 2020
Ph.D. thesis — Network Motif Discovery: Efficiency, Scalability, and Benchmarking, HKU 2024. Full list on Google Scholar.
Projects
Selected systems & experiments
Open source
MOSER
Scalable network motif discovery using serial testing — the core system from my Ph.D., with open-source code.
C++Python
View on GitHub →
Open source
BEACON
A benchmark for efficient and accurate subgraph counting, built to stress-test GNN and algorithmic approaches.
PyTorchGNN
View on GitHub →
Industry
Graph RAG in GaussDB
Associative multi-hop retrieval designed for a production database engine, with a filed patent and product integration.
PythonC++Retrieval
Industry
ContextSniper
Hierarchical code memory and context gating for coding agents — substantial token savings at comparable resolution rates.
LLMMemory
Research
Edge–Cloud Agent Memory
Autonomous memory systems for agents operating across resource-constrained edge devices and cloud backends.
AgentsSystems
Independent
Quantitative Markets Research
Fine-tuning time-series foundation models for BTC forecasting, systematic DCA and grid execution, and options-implied risk-neutral densities.
PyTorchChronosOptions
Teaching & Service
In the classroom
- 2020–2022 Age of Big Data — Tutor & TA, HKU (Prof. Reynold C.K. Cheng)
- 2019–2020 Operating Systems — Head TA, Sharif (Dr. Hamed Farbeh)
- 2019–2020 Advanced Programming — Tutor & TA, Sharif (Dr. Isazadeh)
- 2018–2019 Software Development & Design (Agile) — Tutor & TA, Sharif (Dr. Masoumeh Taromirad)
- 2017–2019 Computer Architecture — Head TA & Tutor, Sharif (Dr. Hamed Farbeh)
Honors & Awards
Recognition
- 2026 FutureStars Medal of Honor — Huawei
- 2020–2024 Belt & Road Scholarship — HKU
- 2020–2024 Postgraduate Scholarship (PGS) — HKU
- 2020 Lee Shau Kee Postgraduate Fellowship — HKU
- 2018 MakeAthon, 3rd runner-up — Sharif University of Technology
- 2013 Tehran Open Chess Championship, 2nd runner-up
- 2012 Member, National Organization for Development of Exceptional Talents (NODET)
- 2010 Champion, Hong Kong Inter-School Chess Championship
Beyond the lab
The rest of it
Competitive Chess
A tournament player since school. Champion of the Hong Kong Inter-School Chess Championship and 2nd runner-up at the Tehran Open. Chess is still how I think about search, evaluation, and knowing when to stop calculating.
Markets & Quant Research
An ongoing independent research interest: fine-tuning time-series foundation models for crypto forecasting, building systematic DCA and grid execution, and recovering risk-neutral densities from options data.
Languages
- PersianNative
- EnglishFluent
- ChineseBasic
- ArabicBasic
Stack
Tools I reach for
Languages
- Python
- C++ · C
- Rust
- SQL
- Java · JavaScript
- Bash · MATLAB
ML & Data
- PyTorch
- TensorFlow
- scikit-learn
- NumPy · pandas
- Jupyter
Databases
- GaussDB
- Neo4j
- PostgreSQL · MySQL
- MongoDB · Redis
Infrastructure
- Git · CI/CD
- Docker · Kubernetes
- Jenkins · Ansible
- Prometheus · Grafana
- AWS · Azure
Contact
Let’s talk systems, models, or markets
Open to research collaborations, technical conversations, and opportunities across databases, machine learning, and production ML systems.