Agentic AI & MLOps / 3D.S.F
AgenticIA · Multi-agent ML / LLM automation
A multi-agent platform connecting ingestion, training, evaluation, deployment and monitoring.
- My role
- Internship · Agentic AI & MLOps
- Period
- Jan — May 2026
Inside the system
Connect the model lifecycle
A conceptual overview of the AgenticIA workflow described in my CV.
- 01
Ingest
Prepare and version data
- 02
Train & tune
Experiments and HPO
- 03
Evaluate & deploy
Model evaluation and serving
- 04
Monitor
Performance and drift
Orchestration
LangGraph coordinates the workflow
Feedback
Diagnose → corrective action → retrain
01 / The context
What the project addresses
ML and LLM workflows involve many connected steps and tools. AgenticIA brings them together through a multi-agent automation platform.
02 / My contribution
What I worked on
Designed and developed the platform, integrated lifecycle tools and implemented an adaptive monitoring and corrective-action loop.
The challenge
Coordinate the model lifecycle and connect performance monitoring to actions such as diagnosis and retraining.
03 / The approach
How the pieces fit together
- Orchestrate ingestion, training, hyperparameter optimization and evaluation with LangGraph.
- Integrate MLflow and DVC for experiment tracking and versioning, and Optuna / Ray Tune for optimization.
- Connect FastAPI, Docker and Kubernetes with deployment workflows, plus vLLM and LangSmith for LLM workflows.
- Use monitoring and drift detection to let agents diagnose degradation and trigger corrective actions.
04 / The outcome
What the work delivered
- Connected ML / LLM lifecycle stages in one agentic workflow.
- Implemented a feedback loop for monitoring, diagnosis and corrective action.