Mahdi Davari, Phd Email & Phone Number

New York, Ny, Us

New York, Ny, Us

New York, New York, Us

• Created multiple ML/DL pipelines in Python, RESTful API with Flask, webapps with Dash. Containerized 8 services (ML models + API) with Docker, Jenkins, OpenShift and deployed them to Citi OpenShift container platform.• Trained multiple computer vision models for end-to-end table detection and table content extraction incorporating pattern recognition on images. Leveraged Cascade mask RCNN (a SOTA multi-stage architecture, with HRNet backbone), using PyTorch, MMCV, MMDet, TorchVision, ONNX, transfer learning from OpenTable dataset and fine tuned on custom datasets, ended up in detection accuracy of >97% .• Created multiple ML services ranging from Named Entity Recognition (NER) for determining labels/tags of contents/entities, e.g., in unstructured documents, to CRF based address parsing, initially trained on OpoenSteetMap and OpenAdresses and fine tuned as per the customer requirements. Used SpaCy, TensorFlow, DeepPavlov, Libpostal, Typer and Flask.• Created multiple end-to-end pipelines (from data pipeline to deploy pipeline through Jenkins). Deployed variety of ML services in OpenShift clusters. Created multiple python based custom Machine Leaning frameworks and packaged them using Python Package Index standards and deployed to client’s PyPi repo.

Bangalore, Karnataka, In

• Led a computer vision project for GUI element extraction and managed a team of data scientists. A patent filing has been initiated for our end-to-end solution.• Provided continues improvements for the application release and deployment management process, workflow and post implementing monitoring and evaluation.• Developed a user activity tracking model using traditional computer vision in addition to the state-of-the-art deep learning models (used Keras, OpenCV, CI/CD with GitLab & Jenkins pipeline)

Bangalore, Karnataka, In

• Business use case implementation of graph data science/graph database with Neo4j (Cypher) andMarkLogic (XQuery) and demonstrated performance, reliability, and integrity of the flexible data modelwith connected data.• Pilot implementation of recommender systems e.g., collaborative filtering (SVD, SVD++, ALS, NMF), content based (KNN, TF-IDF), context aware, session based, hybrid methods (LightFM), etc.• POC on text summarization with implementing a reusable pipeline for abstractive and extractive text summarization approaches, leveraging BERT, T5, and BART NLP models.• POV on distributed learning/big data technologies, covering variety of big data platforms, and distributed learning frameworks (Spark, DASK, RAPIDS (CuDF, CuML), XGBoost4j-Spark, MMLSpark, Koalas, Polars, etc.• Created multiple reusable pipelines in Python, RESTful APIs with Flask, interface with OpenAPI (Swagger), webapps with Dash.• Collaborated effectively within a small team environment as well as work tasks independently

Stony Brook, Ny, Us

• Designed and implemented the Finite Pressure Temperature Elasticity (FPTE) package with more than 2000 lines of code used by researchers worldwide (available on my GitHub)• Helped PI recognize over $300,000 in grant funding by skillfully leading a team of scientists to complete all projects on time, resulting in grant renewal every year.Theoretical simulations based on quantum mechanics have played an increasingly important role in Earth sciences. Therefore, we apply the most advanced and recently developed tools in computational physics to address problems related to the mantle, core and core-mantle boundary region.My research focus is on materials of the most enigmatic regions of the Earth which will be studied using advanced quantum-mechanical simulations. Central to this study is our in-house developed evolutionary algorithm USPEX, interfaced with first-principles electronic structure calculations. This approach has produced many important predictions in mineralogy and physics, successfully confirmed by subsequent experiments.

Stony Brook, New York, Us

My primary research focus is to predict novel materials for energy storage, delivery, and technological applications. My research encompasses materials design, crystal structure prediction, superconducting properties calculation using ab initio quantum mechanical simulations. The goal of my work is to guide experimentalists towards achieving the high-Tc superconductivity.

Stony Brook, Ny, Us

• Co-authored a book chapter, published 15+ peer reviewed journals including a Nature Communication article, 4 invited/contributed talks at conferences.• Designed and modeled new inorganic materials with desired properties using ab initio methods.• Predicted novel compounds under high pressure conditions (borides, hydrides, oxynitrides, etc).My research will address the needs of future energy systems by discovering novel materials for energy storage, delivery, and technological applications. The main aspect of my research is to develop a new methodology for the prediction of novel polyhydride materials with key properties specifically enhanced for application to future energy technologies. By combining theoretical predictions and experiments at various thermodynamic and kinetic conditions, a targeted material design strategy will be developed. This will be realized through scientific discoveries of novel materials and demonstrations of successful synthesis of polyhydrides with enhanced superconducting and hydrogen storage properties.

Stony Brook, Ny, Us

• Designed and developed crystal structure prediction algorithms, and integrated it in USPEX code (used by over 4000 researchers worldwide)• Implemented a descriptor of structure (fingerprint function), capable of the quantification of energy landscapes of 2D materials• Developed a method for two-parameter optimizations, enabled the search for materials with desired properties along with the stability

• Designed, developed and managed USPEX website and user data on SQL databases• Configured and managed high performance computing cluster on Red Hat/Debian platform• Collaborated effectively within a small team environment as well as work tasks independently

Doctor Of Philosophy (Ph.D.), Computational Phys/Chem

Data Science Career Track

Master'S Degree, Condensed Matter And Materials Physics

Bachelor'S Degree, Condensed Matter Physics