< Back to Home
// SYSTEM DOSSIER

Curriculum Vitae

DOWNLOAD STATIC CV //

01 // CORE ARCHITECTURES & SKILLS

ML/DL

  • â–ªPyTorch
  • â–ªTensorFlow/Keras
  • â–ªScikit-Learn
  • â–ªLLMs
  • â–ªHugging Face
  • â–ªTransformers
  • â–ªOpenAI Agents
  • â–ªLangChain
  • â–ªCrewAI
  • â–ªNLP
  • â–ªRAG

Programming

  • â–ªPython
  • â–ªSolidity
  • â–ªSelenium
  • â–ªBeautifulSoup

databases

  • â–ªPostgreSQL
  • â–ªSQL Server
  • â–ªMongoDB
  • â–ªFAISS
  • â–ªChroma

tools_and_cloud

  • â–ªGitHub
  • â–ªDocker
  • â–ªFastAPI

02 // ENGINEERING HISTORY & CHRONOLOGY

DEC 2025 — PRESENT

AI Engineer // Simiaroom

Matchmaking SystemsPrompt EngineeringLearning-to-RankPostgreSQLMongoDBFastAPIDockerFeature Engineering
  • >Machine Learning Architecture: Architected a high-throughput, AI-driven therapist matchmaking engine utilizing a complex 6-layer processing pipeline to achieve precise patient-therapist alignment.
  • >LLM Clinical Alignment: Engineered an LLM-based clinical translation framework to normalize unstructured patient intake data into high-dimensional semantic features.
  • >Learning-to-Rank Optimization: Deployed an XGBoost Pairwise Ranker model to optimize recommendation weights, significantly enhancing matching accuracy and accelerating successful patient onboarding.
JUL 2025 — PRESENT

Applied AI Engineer // SB Immigration Services

RAGLLMsOpenAI WhisperMongoDB Atlas Vector SearchPersian EmbeddingsSemantic SearchAWS SageMaker
  • >Intelligent RAG Architecture: Engineered a high-fidelity Retrieval-Augmented Generation (RAG) assistant utilizing OpenAI LLMs and advanced prompt engineering to automate and streamline context-aware answers for complex immigration workflows.
  • >High-Dimensional Vector Ingestion: Built localized, cross-lingual semantic indexing pathways by embedding raw text transcriptions into a MongoDB Atlas Vector Store using Jina-embeddings and targeted Persian embedding models to maximize retrieval recall.
  • >Multimodal ETL Pipeline: Developed an automated data harvesting pipeline that ingests social media video content, programmatically extracts core audio tracks, and processes automated transcription loops via OpenAI Whisper pipelines hosted natively on AWS SageMaker.
JUL 2023 — JUN 2025

ML Engineer and Team leader // Mass Group

  • >Transformer Optimization: Fine-tuned a compact BERT architecture (`bert_en_uncased_L-10_H-128_A-2`) for text classification on real-time data streams, securing a competitive validation score of 0.797.
  • >Parameter-Efficient Fine-Tuning (PEFT): Enhanced a T5-small conditional generator on the IMDB dataset using advanced optimization frameworks, including LoRA (Low-Rank Adaptation), Adapters, and Soft Prompt Tuning to minimize training compute overhead.
  • >Algorithmic Asset Forecasting: Built a time-series machine learning model tracking 1-hour and 4-hour candlestick intervals to forecast Bitcoin valuation movements, achieving a reliable 70% directional trend accuracy.
  • >Multimodal Sentiment Integration: Scaled the cryptocurrency predictive engine by integrating an NLP pipeline designed to parse macro news sentiment and shift trend thresholds based on real-time external events.
  • >Computer Vision & OCR: Fine-tuned a ResNet-34 backbone to construct a highly performant image-to-text conversion model, drastically improving machine interpretation of diverse handwritten inputs.
  • >Deep Learning Classification: Architected and deployed a custom predictive classification model in PyTorch to automate risk profiling and evaluate consumer travel insurance purchase propensities.
AUG 2022 — JUN 2023

ML Engineer | Internship // MAPSA HR

  • >Neural Machine Translation: Built an English-to-Spanish translation pipeline from scratch, developing a custom Sequence-to-Sequence network with Attention mechanics in PyTorch alongside `NN.TRANSFORMER` and `torchtext` for deep language modeling.
  • >Computer Vision Optimization: Designed and deployed a real-time mask detection system leveraging an EfficientNetB0 architecture combined with advanced deep learning and image-processing techniques.
  • >Semantic Recommendation Engine: Developed a sophisticated lyric-based song recommendation system utilizing `sentence_transformers` for embedding vectorization and Scikit-learn Cosine Similarity algorithms for feature mapping.
  • >Predictive Regression Modeling: Engineered a robust real estate valuation model for the Tehran housing market, optimizing feature transformations to achieve a reliable 89% prediction accuracy.
  • >Object Detection Deployment: Implemented and evaluated a high-throughput YOLO object detection network trained on the benchmark COCO dataset for real-time multi-class feature extraction.
  • >Time-Series Forecasting: Applied advanced statistical time-series analysis and regression techniques to a bicycle mobility dataset to model historical demand and predict usage trends.
  • >Statistical Data Analysis: Conducted comprehensive exploratory data analysis on the classic Titanic dataset utilizing NumPy, Pandas, and SciPy, generating high-fidelity multi-variable distribution plots using Matplotlib and Seaborn.
  • >Large-Scale Data Auditing: Maintained, cleaned, and analyzed a massive public safety dataset containing King County Health Department inspection logs from 2006 to present day, isolating key operational health anomalies.