My Work
🖼️ Automatic Image Caption Generation
Computer Vision & NLP | Dataset: Flickr8k.
This project explores image captioning, a challenging multimodal task in artificial intelligence that consists of generating a coherent textual description for a given image. It combines deep learning techniques from both computer vision and natural language processing.
🎯 Objectives
- Generate meaningful captions for images.
- Implement a model inspired by Show and Tell.
- Combine visual and semantic features.
🧠 Model & Architecture
- Image: ResNet pretrained on ImageNet; feature vectors from the penultimate layer.
- Text: Word embeddings with GloVe; vocab of 1660 words in a (1660, 220) matrix.
- Captures semantic word relations via vector arithmetic.
🩺 Machine Learning Model API with Streamlit Interface
Diabetes prediction deployed on Google Cloud Run using Artifact Registry.
A diabetes prediction model wrapped in a FastAPI REST API with a Streamlit frontend. The app is dockerized for easy deployment and hosted serverlessly.
🎯 Features
- Accurate diabetes risk prediction.
- FastAPI REST API for model serving.
- Interactive Streamlit user interface.
- Docker container for portability.
- Deployment on Google Cloud Run.
🧪 Testing
- Local tests ensured API and UI reliability before deployment.
🔗 Technologies
- FastAPI, Streamlit, Docker
- Google Cloud Run, Artifact Registry
🧠 MySQL Query Generator using LangChain
This project uses LangChain and LLMs to generate intelligent SQL queries from natural language inputs. It dynamically identifies the right database tables and formulates executable queries.
🎯 Key Features
- Dynamic table detection based on user queries.
- SQL query generation using large language models (LLMs).
- Contextual memory for handling follow-up questions.
- Answer rephrasing and optimized SQL output.
🛠️ Technologies
- LangChain & OpenAI
- Pandas
- Jupyter Notebook
📍 Marker Detection and Classification
Tool for professionals to mark zones of interest that may present anomalies. This project focuses on extracting and classifying these markers to assist expert decision-making.
🎯 Objectives
- Identify pointed zones and apply various image processing techniques.
- Determine arrow orientation (direction and sense).
- Define extraction zone for the region of interest (ROI).
- Apply multiple image treatments on the ROI.
📊 Data Professionals Trends Analysis – Power BI
Business Intelligence | LinkedIn Job Market Data
This Power BI project explores data-related job trends by building an interactive dashboard from LinkedIn data. It highlights job titles, in-demand skills, and geographic distribution.
🎯 Objectives
- Visualize hiring trends for data jobs using an intuitive dashboard.
- Identify key skills and most common job titles across locations.
- Present insights clearly through clean and dynamic visualizations.
🛠️ Tools & Techniques
- Power Query: for data cleaning and transformation.
- DAX Measures: for creating custom KPIs.
- Interactive dashboard built in Power BI.
- Includes .pbix file, Excel source data, and PDF report.
About Me
I'm Aissatou Signate, a passionate Machine Learning Engineer and Data Scientist based in Île-de-France, France. I specialize in building smart, scalable AI solutions, particularly in computer vision and deep learning.
With a dual background in computer science and economics, I bring both technical expertise and analytical thinking to solve complex problems.
Over the past few years, I’ve worked on various impactful projects—from medical image analysis to data-driven decision-making tools. Whether it’s fine-tuning neural networks or exploring new datasets, I thrive on turning data into insights and innovation.
📍 Based in Paris region, currently open to V.I.E missions, full-time roles, or AI consulting projects.
Beyond coding, I’m passionate about ethical AI, continuous learning, and creating tools that make a real difference.
CV available upon request
