Mamadou-dian Djalo

The project aims to put artificial intelligence at the service of recruiting teams by transforming raw documents—such as resumes and job descriptions—into structured, instantly usable data. It consists of creating an automated Recruiting CRM with workflows that can extract key skills, summarize experiences, and generate a target profile perfectly aligned with the role’s requirements. A matching engine then analyzes this profile against the internal talent pool, taking into account contextual parameters such as seniority, location, and candidates’ strengths, to deliver reasoned recommendations. The interface centralizes all searches, offers relevant filters, and allows real-time tracking of processing status. This setup accelerates the prioritization of top profiles and optimizes productivity across the entire sourcing process. Tasks : • Automate secure ingestion of resumes and job descriptions: upload portal, AI inspection on arrival, Celery queueing to trigger GPT-5 processing as soon as files are received. • Design extraction pipelines: Jinja prompts orchestrating GPT-5 to structure experiences, skills, and business requirements into standardized Pydantic schemas. • Develop a hybrid matching engine: 1,536-d embeddings (OpenAI) + pgvector (cosine) vector search, explainable ranking that combines similarity, business weighting, and strict validation of must-have skills. • Set up event-driven APIs: trigger workflows (ingestion, parsing, search, resume updates), real-time event tracking, and history of recommendations. • Orchestrate & monitor async processing: Celery/Redis supervision, exponential retries, full traceability, SLOs (latency, throughput, success rate). • Model Talent data: candidate/experience/skill repository optimized for semantic search and operational filters (location, seniority, availability). • Build the recruiter interface (Next.js): smart forms, advanced filters, server-side pagination, real-time toasts, local persistence to keep AI analysis context. • Deliver enriched search history: visualize matches, one-click access to profiles, score explanations highlighting presence/absence of must-have skills to speed up decisions. • Industrialize the environment: Docker packaging (API, workers, database, proxy) for reproducible deployments and controlled scaling. • Support recruiting/ops teams: scoping workshops, regular demos of AI capabilities, training on score interpretation, and ongoing post-go-live support. Tech Stack • Backend/Data: Python (FastAPI, Celery, SQLAlchemy), PostgreSQL + pgvector, Supabase Storage, Redis, OpenAI (GPT-5 reasoning, embeddings 1 536-d). • Frontend: TypeScript, Next.js (App Router), React Query, Tailwind CSS, Sonner. • Ops & Observabilité : Docker / Docker Compose, logs & métriques centralisés (latence, throughput, succès/erreurs).

Tasks: - Develop agentic workflows using LangChain. - Monitor and analyze the cost, latency, and quality of responses provided by LLMs using LangFuse. - Deploy PostgreSQL vector databases. - Strengthen the robustness of LLM responses using Instructor. - Develop file ingestion pipelines with verification systems using Cloud Storage and Cloud Functions. - Manage the lifecycle of files in Cloud Storage. - Profile BigQuery tables to detect and categorize sensitive data using Data Loss Prevention (DLP). - Administer access to BigQuery tables with IAM. - Implement cryptographic anonymization using the DLP API and BigQuery Column-Level Security. Technical Environment: - Python, Vertex AI, Cloud Run, LangChain, LangFuse, Cloud Functions, Pydantic, GitHub Actions, Terraform, Cloud Storage, BigQuery, PostgreSQL, Gemini, OpenAI, Claude, Data Loss Prevention, Pub/Sub, IAM.

Task: - Develop file ingestion pipelines using BigQuery and Cloud Storage. - Build data transformation workflows with Cloud Workflows and SQL. - Implement dynamic triggers for transformation workflows and analytical tables ingestion. - Orchestrate and monitor workflows. - Provision and manage cloud infrastructure on GCP with Terraform. - Develop and continuously deploy with Terraform, GitHub Actions, and Cloud Build. - Optimize SQL queries and table structures to reduce costs. - Model data to improve dashboard performance in Power BI. Technical environment: - BigQuery, Cloud Run, Workflows, PubSub, Event Arc, CloudBuild, Secret Manager, Terraform, Python, SQL, Cloud Storage, Cloud Function.

Task: - Develop data transformation pipelines using Presto SQL. - Build an accessory recommendation system using collaborative filtering and content-based filtering. - Develop a gradient boosting model to predict vehicle purchase probability. - Implement an MLOps infrastructure to automate model training with Databricks and dbx. - Develop DAG workflows to create features from web behavior and customer data using PySpark. - Perform data exploration and analysis with Python, Pandas, and Plotly. Technical environment: - BigQuery, Terraform, Github Actions, Mlflow, Databricks, Azure Data Lake, CDP Treasure Data, Python, Spark, Pandas.

Tasks: - Develop and maintain Python-based applications using Django for both back-end and front-end solutions. - Design and implement RESTful APIs to enable seamless integration with front-end applications and other services. - Build dynamic templates using Jinja to enhance the user interface and interaction. - Perform data wrangling and preprocessing tasks on structured and unstructured data to prepare it for analysis or further processing. - Design and implement algorithms to map, transform, and clean data from YAML files into structured formats suitable for applications or analysis. - Write comprehensive unit tests and integration tests to ensure the reliability, scalability, and performance of Python applications. - Deploy and maintain Python-based applications in cloud environments, ensuring high availability and performance. Technical environment: - Python, Django, Jinja, REST API, Data wrangling, YAML, Algorithm development, Unit testing, Git, Docker, Cloud deployment.