Maryam A.

AI Engineer / Data Scientist

Prague, Czech Republic

Experience

Sep 2025 - Present
4 months

AI Engineer / Data Scientist

Space Physics Group, Charles University

  • Using generative AI techniques to accelerate and enhance Coronal Mass Ejection (CME) simulations for improved space weather prediction.
  • Mentored junior researchers in ML model development, data preprocessing, and model validation.
Jan 2025 - Dec 2025
1 year

Digital Twin Floorplan Generation (Python)

Self-directed

  • Processed 3D LIDAR point clouds with RANSAC and DBSCAN to generate accurate 2D floorplans.
Jan 2025 - Dec 2025
1 year

Transformer-Based Semantic Embedding Prototype (Python)

Self-directed

  • Prototyped semantic search using Sentence Transformers and FAISS vector similarity to explore domain-specific retrieval.
  • Compared cosine similarity and hybrid search strategies for identifying context-relevant plasma events.
  • Gained practical understanding of embeddings, vector stores, query expansion, and RAG workflows.
Feb 2024 - May 2024
4 months
Taiwan, Province of China

Visiting Researcher (HPC AI Workflows)

National Center for High-Performance Computing (NCHC)

  • Automated data ingestion and analysis pipelines in HPC environments using Python and Linux-based scripting.
Jan 2024 - Dec 2025
2 years

Region Identification in Spacecraft Data using Machine Learning (Python)

Self-directed

  • Processed THEMIS spacecraft time series data, performed feature extraction, and trained supervised ML models.
  • Developed Random Forest and Neural Network classifiers using scikit-learn and TensorFlow to automate region identification.
Jan 2022 - Dec 2023
2 years

Magnetopause Location Modeling with ANN (IDL)

Self-directed

  • Trained an ANN to predict boundary locations using solar wind parameters, IMF orientation, and corrected Dst.
Oct 2021 - Sep 2025
4 years
Prague, Czech Republic

Research Assistant

Charles University

  • Conducted AI-driven analysis of plasma environments using spacecraft data from THEMIS, Geotail, and Magion 4 missions.
  • Built predictive models to estimate magnetopause positions and classify solar wind regions.
  • Developed end-to-end data pipelines for feature extraction, normalization, and model training using Python and TensorFlow.
  • Applied machine learning models (neural networks, random forests) to classify complex satellite datasets.
Dec 2020 - Aug 2021
9 months
Iran, Islamic Republic of

Telecom Pre-Deployment Engineer

Borje Noor

  • Designed FTTx broadband network topologies and optimized component selection using COMSof FTTx.
  • Supported planning for 5G infrastructure and coordinated with engineering and procurement teams.
Jul 2020 - Dec 2020
6 months
Netherlands
Remote

Remote Guest Researcher (Quantum Systems)

QuTech, TU Delft

  • Theoretically analyzed two-qubit photonic crystal systems and explored entanglement mechanisms in solid-state photonics.
Oct 2016 - Jun 2020
3 years 9 months
Iran, Islamic Republic of

Research Assistant, Photonics Group

Shahrekord University

  • Simulated electromagnetic and photonic systems using COMSOL and Lumerical for sensing and energy harvesting applications.
  • Designed and optimized plasmonic waveguides, resonators, and solar cells.

Summary

AI Engineer with a PhD in space physics and a strong background in applied machine learning, data pipelines, and scientific computing. Experienced in developing classification models using neural networks and decision trees for large-scale, domain-specific datasets. Knowledgeable in modern NLP techniques including transformer-based models (e.g., BERT, GPT), vector similarity search, and retrieval-augmented generation (RAG). Strong Python programmer with expertise in TensorFlow, scikit-learn, and data visualization. Actively exploring LLM-based architectures and semantic search integration for specialized domains.

Languages

English
Advanced
Czech
Elementary

Education

Lorem ipsum dolor sit amet

PhD · space physics

Certifications & licenses

Honored Certificate

Need a freelancer? Find your match in seconds.
Try FRATCH GPT
More actions