Mohamed Camara

Develop an 'AI Profile Agent' solution using agents hosted in Github Copilot such as Claude Sonnet, by using the Ollama local server and large language models (LLMs). Design a technical, backend, and pipeline architecture for the AI agent solution (based on a large language model across all structured and unstructured documents from SharePoint, from ingestion to the complete solution (Webhook, SharePoint Rest API, OCR, LLM, Azure AD...). Develop the 'AI Mandate Agent' solution using Llama, Qwen, or Mistral type models with optimized choices based on the complexity levels of LLM prompts. Extract requirements, experiences, prerequisites, and technologies with a hybrid method (regex+LLM+Fallback) to clearly identify the CV profiles sought. Use LLM models for selecting and sorting the most relevant documents in SharePoint that will serve as input mandate data corpus. Establish the matching of Profilo CVs and mandates carried out that are most correlated with the requirements of tenders and intervention requests. Establish a classification and ranking using a hybrid weighted method (semantic similarity search + keywords + technological density of the text).

Build consumption tracking dashboards using Quicksight and develop the FINOPS Management Platform (PGF) solution. Design a consumption tracking tab, daily and monthly cumulative consumption. Develop a tool for tracking discrepancies between commitment amounts by contract and actual expenses according to the organization and client. Establish tracking of key performance indicators (KPIs) focused on the application solution and comparisons from the beginning to the end of the period over a rolling N-month period. Track discrepancies between dollar amounts from Focus files integrated into the platform and invoice amounts with a customizable discrepancy threshold highlighted by visual alerts. Implement a Python script integrating Quicksight export API functionalities into a Lambda function for Quicksight analysis backups. Develop a backup architecture for Lab, Dev, and Production environments. Write documentation in Azure DevOps describing the content of the different dashboard tabs and their specifications.

Build a knowledge base from tickets and errors recorded on JIRA and extract data in a structured way using the Mistral generative AI tool. Develop and implement an artificial intelligence tool for predicting the occurrence of errors identified in the knowledge base using Expert System type models for predictive maintenance of errors that may occur based on the association of technologies. Build a dashboard on AWS Quicksight with a Supervisor view including call status indicators. Build a dashboard with an Agent view for monitoring individual performance in terms of satisfaction rate, response speed, waiting time, or call duration indicators. Build a dashboard with an Administrator view for monitoring KPIs and SLA indicators, call duration metrics (AHT, ACW...) by day of the week and time of day. Develop a chatbot prototype composed of Amazon Lex, Amazon Bedrock, and AWS Lambda technologies for their orchestration. Build a Bedrock Knowledge base to provide contextual information from data sources. Configure and set up Amazon Lex to serve as an exchange interface between the user and AI, test and calibrate the parameters of the selected Claude Anthropic and Titan models based on the responses provided in collaboration with the client's internal teams to validate their performance and relevance.