Abdelaziz B.

• - Designed agentic LLM flows (qualification, matching, resume gen/checks) with prompt engineering, embeddings, caching, and strict validation.

• - Led a production-grade parsing/intelligence platform for hiring workflows.

• - Built modular FastAPI services (docs/resumes/emails/jobs/timesheets) with Pydantic and repos.

• - Implemented hybrid search: Elasticsearch filters + embedding vectors for high-precision retrieval.

• - Worked with private and open-source LLMs (OpenAI, Gemini, Claude, Llama), tuning prompts for accurate answers.

• - Shipped AWS SQS pipeline (DLQ, long polling, visibility timeouts) with resilient workers and Supervisor-based concurrency.

• - Implemented reliable webhooks with retries and standardized payloads.

• - Deployed Dockerized services on AWS ECS/Copilot with SSM secrets and observability.

:–- Created a document digitization solution for visually impaired users, researchers, and professionals.

• - Utilized TensorFlow with EfficientNet for accurate page-flip detection and image classification.

• - Integrated PaddleOCR to extract text with high precision and maintain formatting.

• - Built a user-friendly Flask interface, incorporated OpenAI APIs for text-to speech, and deployed the application using Docker, AWS via Terraform, and automated CI/CD pipelines with GitHub Actions. : Python, TensorFlow, PaddleOCR, Flask, Docker, Terraform, AWS, GitHub Actions, OpenAI API. :

• - Applied NLP techniques to accelerate candidate selection by using models such as BERT, GloVe, NER, and Doc2Vec to calculate cosine similarity between job titles and search keywords.

• - Cleaned and normalized data for robust analysis and re-ranked candidates using RankNet when starred by recruiters.

• - Developed an LLM-based application with LangChain for enhanced candidate matching and a UI for loading data and choosing between local (Ollama) or cloud-based models. GitHub: Potential Talents :

• - Designed classification models to identify prime customer segments, optimizing marketing efforts and cutting over 2000 hours of phone calls. : Pandas, Numpy, Matplotlib, Seaborn, Pycaret, Plotly, Hyperopt, Optuna, TPOT. :

• - Developed a model predicting customer satisfaction with 87% accuracy by employing feature importance and elimination (RFE). : Pandas, Numpy, LLM, LangChain, gensim, matplotlib, NLTK, Scikit-learn, Streamlit.

• - Team member of Reinforcements Learning Human Feedback (RLHF) for improving ChatGPT python code generation answers.

• - Responsible for validating and ranking multiple ChatGPT AI model's responses towards user's queries.

• - Wrote high-quality and optimal code in Python to solve the puzzle tasks to feed the AI models.

• - Provided necessary and clear explanations of the thought process/logic while writing the code in Jupyter notebooks.

• - Worked with problem-solving to handle complex situations with ease.

• - Debugged and created necessary documentation.

• - Provided feedback while doing peer review of code.