CV Amazon Kinesis : Trouvez le meilleur profil indépendant

Exemple d'expériences d'Ismail,
freelance AMAZON KINESIS habitant ?

WORK EXPERIENCE

Freelancer Oct. 2022 - present
Researcher Oct. 2022 - present
Mohammed VI Polytechnic University, Ben Guerir
- Build and deploy models trained on medical datasets, and develop new imputation techniques to ensure data consistency in the medical field.

Data Scientist,Intern Mar. 2022 - Oct. 2022
EVEAD GROUP, Csablanca
- Creation of a classification container and extraction of a maximum of data from each image, small or large in the
advertisements.

Web Developer,Intern Sep. 2019 - Dec. 2019
ECS informatique, Rabat
- Developed a website for an investment company in Cultural Industries.

PROJECTS

Enhanced Toxicity Mitigation in Summarization Models Using FLAN-T5 with Reinforcement Learning and
PEFT Jan. 2024 - Mars. 2024
Self employed
- Implemented advanced techniques to fine-tune the FLAN-T5 model, focusing on reducing toxic content in summaries
using Meta AI’s hate speech reward model and Proximal Policy Optimization (PPO).
-Tools: FLAN-T5, PPO, Meta AI’s hate speech reward model, Google Cloud Platform (GCP).

Generative AI for Dialogue Summarization Using Prompt Engineering Sep. 2023 - Nov. 2023
Self employed
- Applied generative AI techniques to improve dialogue summarization by exploring the impact of various prompt
engineering strategies on model performance.
-Conducted summarization tasks with and without prompt engineering, including zero shot, one shot, and few shot
inferences.
-Compared instruction-based prompt approaches using FLAN-T5 and refined generative outputs by adjusting prompt
configurations and evaluating their effectiveness.
-Tools: FLAN-T5, Google Cloud Platform (GCP).

Fine-Tuning FLAN-T5 for Enhanced Dialogue Summarization Using Parameter Efficient Fine-Tuning (PEFT)
Jun. 2023 - Aug. 2023
Self employed
- Fine-tuning the FLAN-T5 model from Hugging Face for improved dialogue summarization, applying both full finetuning and Parameter Efficient Fine-Tuning (PEFT) techniques. Evaluated model performance using ROUGE metrics.
-Prepossessing the dialogue-summary dataset, fine-tuning the FLAN-T5 model, and conducted both qualitative and
quantitative evaluations using ROUGE metrics.
-Configuring PEFT/LoRA adapters, training the PEFT model, and assessing its performance qualitatively and quantitatively, comparing it to full fine-tuning to highlight the trade-offs between performance and efficiency.
-Tools: Hugging Face Transformers, FLAN-T5, Google Cloud Platform (GCP), Python, ROUGE Metrics.

End-to-End TFX Pipeline for Evaluating Census Income Classification Models Feb. 2023 - Mars. 2023
Self employed
- Building a TFX pipeline that integrates various components including ExampleGen, StatisticsGen, SchemaGen,
ExampleValidator, Transform, Trainer, and Evaluator.
-Managing dataset ingestion and preprocessing, including scaling numeric features and one-hot encoding categorical
variables.
-Developing and trained a classification model using a wide and deep model architecture, with provisions for tuning
and performance optimization.
-Configuring and executed the Evaluator component to compute performance metrics, validate model improvements,
and compare new models with previous versions.
-Utilizing TFMA to assess model performance against defined metrics and thresholds, ensuring the model meets
quality standards before deployment.
-Tools: TensorFlow Extended (TFX), TensorFlow Model Analysis (TFMA).

Advanced Data Pipeline Development for Forest Cover Type Prediction Dec 2022 - Jan. 2023
Self employed
- Development of a comprehensive data pipeline solution for predicting forest cover types using the Cover Type dataset,
enhancing the model’s accuracy and efficiency in real-world applications.
-Ingestion of the dataset and generated statistics to understand data distribution.
-Creation of a schema based on domain knowledge and established schema environments.
-Visualization of dataset anomalies and engineered features.
-Tracking the data pipeline provenance using ML Metadata.
-Tools: TFX components, TFDV, ML Metadata.

Developed a Feature Engineering Pipeline using TFX Oct 2022 - Nov. 2022
Self employed
- Built an end-to-end pipeline using TensorFlow Extended (TFX) for feature engineering on the Census Income dataset.
- Implemented components like ExampleGen, StatisticsGen, SchemaGen, ExampleValidator, and Transform for data
ingestion, analysis, validation, and preprocessing.
-Tools: TensorFlow, TFX, TensorFlow Transform (TFT)

A Big Data architecture for real-time Covid-19 data analysis using Amazon Kinesis, Amazon EC2, Apache
Spark, ElasticSearch and Kibana Sep 2021 - Jan. 2022
Mohammed VI Polytechnic University, Ben Guerir
- development of a robust architecture able to analyse in real time the Covid-19 data in order to build a sentiment
analysis dashboard..
-Tools: Amazon Web Services, Spark-Nlp, EC2, Kinesis, Kibana, EMR, Elastic Search.

Disease Prediction Chatbot May 2021 - Sep. 2021
Mohammed VI Polytechnic University, Ben Guerir
- Development of a web application in the form of a conversational agent that simulates a medical consultation.
-Tools: BeautifulSoup,NLP,Scikit-learn,Flask,Bootstrap.. .

Predict the subscription to a banking marketing campaign Mar. 2021 - May 2021
Mohammed VI Polytechnic University, Ben Guerir
- Development of a WEB application based on a system for predicting people likely to subscribe to a banking marketing
campaign.
- Tools: Scikit-learn, Flask.

Simulation of the spread of COVID-19 Dec. 2020 - Feb. 2021
Mohammed VI Polytechnic University, Ben Guerir
- Simulation of the spread of COVID-19 in a restricted area using agent-based modeling.
- Tools: Python, Mesa.

Online shop Apr. 2020 - May. 2020
Faculty Of Science, Rabat,Morocco
- Develope an online store.
- Tools: Bootstrap , Mysql ,PHP , HTML5 , CSS , JavaScript.

Data Leak Prevention Application Sep. 2019 - Jan. 2020
Faculty Of Science, Rabat,Morocco
- Application that allows the user to have the choice on each file transfer from his device .
- Tools: Java, IHM, Unix, AndroidStudio.

Creation of an Educational Website Dec. 2018 - Jan. 2019
Faculty Of Science, Rabat,Morocco
- The website offers an explanation of math operations before offering a series of exercises to solve.
- Tools: HTML, CSS, JavaScript.