David Sarkissian

Senior ML Engineer, I have the opportunity to work on stimulating topics, particularly in ML and Gen AI.

I have notably worked on:

- Optimizing a forecasting model that encountered significant difficulties during the training phase (3 weeks of training) by reducing this training time to a few hours. The code did not utilize the machine's processing power to its full potential.

- Implementing a Tech for Tech solution using Gen AI to facilitate the work of Data Engineers and Product Owners, particularly on in-depth analysis of Data Lineage stored in BigQuery.

- Implementing a POC on a Gen BI topic of the "Talk to my data" type.

Obviously, all these subjects are conducted on GCP (Vertex AI, BigQuery, Cloud Run, etc.)

Worked at Senoee on a very short mission (10 working days) to industrialize the company's key model (FastText by Meta).

Using GCP services (Vertex AI Pipelines, Bigquery, Artifact Registry, and Cloud Run), I was able to industrialize the model in question with a training pipeline that automates data preprocessing, model training, upload to Model Registry, and exposure on a Cloud Run endpoint.

This industrialization also helped to familiarize the company with the most reliable and secure way to deploy a model in production.

POC on customer churn detection with the creation of a multimodal deep learning model (“image” and tabular data). Vectorization of different customer journeys with models like Doc2Vec.

Design and production deployment of a Deep Learning anomaly detection model based on an AutoEncoder architecture.

Personalized Assistant for a business based on RAG. Notably with Neo4j graph database.