CV

Research Interests

Database Systems; Machine Learning for Data Systems; Adaptive and Learned Data Systems; Query Processing and Optimization; Natural-Language Interfaces to Databases; Graph Data Systems

Education

Temple University

Ph.D. in Computer and Information Sciences, Information Systems - GPA: 3.84/4.00
Philadelphia, PA · 2025-Present

Alzahra University

B.S. in Computer Engineering
Tehran, Iran · 2015-2021

• Undergraduate thesis: Persian Text Normalization and Context-Aware Spelling Correction.
• Developed a context-aware approach for detecting and correcting Persian spelling errors, with particular emphasis on homophones and context-dependent word substitutions that cannot be identified through dictionary lookup alone.
• Investigated lexical ambiguity introduced by languages that share the Perso-Arabic writing system, including Arabic and Urdu, identifying borrowed words and valid cross-lingual forms to distinguish them from true spelling errors.

Sharif University of Technology

Exchange Student, Computer Engineering
Tehran, Iran · 2016-2021

• Completed nine semesters of undergraduate Computer Engineering coursework as an exchange student.

Publications

CypherLens: An Interactive Demo for Evaluating and Diagnosing NL-to-Cypher Systems

Yasmin Mohammadi, et al.
VLDB 2026 Demonstration Track. Accepted.

CypherLens: Graph-Native Evaluation, Diagnosis, and Repair for NL-to-Cypher Systems

Yasmin Mohammadi, et al.
Under review, VLDB 2027.

CypherSem: Graph-Native Semantic Error Analysis for NL2Cypher

Yasmin Mohammadi, et al.
Under review, KDD 2027.

Research Projects

Learned Indexes for Graph Databases

Temple University

• Implemented learned indexing techniques originally developed for relational data and adapted them for evaluation on knowledge graph workloads.
• Evaluated their performance across multiple knowledge graph data distributions to characterize how distributional properties affect learned-index accuracy, efficiency, and applicability to graph data.

CypherSem - Semantic Error Analysis for NL-to-Cypher

Temple University

• Developed a three-level, graph-native semantic error taxonomy for NL-to-Cypher spanning four semantic dimensions, 17 categories, and 54 fine-grained error types, and constructed a 3,568-pair human-annotated benchmark.
• Evaluated seven LLMs for fine-grained semantic error detection and semantic correctness judging, uncovering persistent failures on graph-specific errors and systematic benchmark quality issues.

CypherLens - Evaluation, Diagnosis, and Repair for NL-to-Cypher

Temple University

• Designed a graph-native evaluation framework that canonicalizes and aligns Cypher query outputs across aliases, lineage, graph-shaped values, and derived projections, reducing the average gap to human semantic judgments from 25.69% to 1.56% across three Neo4j databases.
• Developed a deterministic multi-agent repair architecture with specialized Diagnoser, Selector, Corrector, and Verifier agents backed by 28 schema-, structure-, execution-, plan-, and intent-aware tools, improving exact match from 24.3% to 68.4% on the evaluated repair set.

CIRA - Continual Experience Learning for Adaptive Data Synthesis

Temple University

• Developing a training-free continual learning framework that converts human and automated feedback into structured, reusable experiences that improve future synthetic-data generation without updating LLM parameters.
• Building an experience lifecycle that compiles feedback into scoped semantic knowledge, then adds, updates, merges, rejects, and selectively retrieves experiences from an evolving memory to guide subsequent agents.
• Evaluating whether experience accumulation improves expert-validated data quality and difficult-query coverage while reducing human supervision, with NL-to-Cypher as the primary testbed and transfer across domains and to NL-to-SQL.

Knowledge Graph Integration and Federation for Earth Science

Temple University

• Exploring automated integration of heterogeneous Earth-science knowledge graphs through schema and ontology alignment, entity resolution, vocabulary harmonization, and provenance-aware conflict resolution.
• Studying the trade-offs between materializing a unified knowledge graph and preserving independent graphs under a federated query architecture that dynamically retrieves and combines information at query time.
• Developing methods for high-precision semantic alignment while preserving source-specific information, provenance, and evolving scientific vocabularies.

KG-Grounded Scientific Information Extraction and Entity Resolution

Temple University

• Developing context-aware methods for extracting scientific entities and concepts from publications and grounding mentions to canonical entities in domain-specific vocabularies and knowledge graphs.
• Exploring knowledge-graph context, ontology structure, and surrounding textual evidence for resolving ambiguous entity mentions, normalizing scientific terminology, and reducing incorrect entity matches.
• Building a semantically grounded indexing pipeline that links extracted concepts to structured scientific knowledge, enabling more precise retrieval across publications, datasets, models, and domain terminology.

Industry Experience

Jambit GmbH

Software Engineer
Yerevan, Armenia · 2022-2025

• Owned end-to-end development of large-scale production systems for Volkswagen Group vehicle-data platforms, working across frontend UI, backend services, database layers, deployment, production support, and debugging for applications used across Europe, the United States, and Canada.
• Developed and maintained a distributed microservice architecture comprising roughly ten cooperating services built with Java and Vert.x, with service coordination through Hazelcast, multithreaded processing, application-level caching, and high-volume database access.
• Improved performance, scalability, and reliability under high request volumes and time-sensitive vehicle data, including cache-refresh logic, concurrent processing, database optimization, production incident diagnosis, and deployment and operations using PostgreSQL, Docker, Kubernetes, and AWS.

Polixis

Java Developer
Yerevan, Armenia · 2021-2022

• Designed and developed a distributed data migration and streaming platform for moving datasets across heterogeneous storage systems, including MongoDB, Elasticsearch, and ScyllaDB.
• Built data-transfer and real-time migration services using Java, Spring Boot, Kafka, Kafka Streams, Kafka Connect, PostgreSQL, Docker, and Kubernetes.
• Managed database changes across the software release lifecycle by identifying and capturing approved changes in development and test environments and promoting them through test, pre-production, and production stages.

Soha Software Group

Java Developer
Tehran, Iran · 2020-2021

• Developed an event-driven, self-hosted push-notification framework for large-scale message delivery without Firebase, focusing on server efficiency, delivery cost, and mobile battery consumption.

DPI (formerly IBM-affiliated operations)

Junior Java Developer
Tehran, Iran · 2019-2020

• Developed and supported Java/Spring Boot microservices for a CRM platform serving approximately 500,000 active users, using IBM MQ, REST APIs, and DB2 with production deployment and incident support.

Technical Skills

• Programming: Python, Java, SQL, Cypher, Bash, C, JavaScript
• Databases & Data Systems: Neo4j, PostgreSQL, MongoDB, ScyllaDB, Elasticsearch, MySQL, MS SQL Server, DB2, Kafka, Kafka Streams, Kafka Connect
• LLM & ML Systems: Transformers, Tool-Using and Multi-Agent Systems, RAG, LLM Evaluation, Retrieval and Memory Systems, PyTorch, TensorFlow, LangChain
• Distributed Systems & Cloud: Spring Boot, Vert.x, Hazelcast, RabbitMQ, REST APIs, Microservices, Docker, Kubernetes, AWS, Jenkins
• Research Methods & Tools: Linux, Git, LaTeX, benchmark and dataset construction, human annotation, experimental evaluation