Looking for a job

Farhi User

@farhiuser

Message

Lead Machine Learning Engineer at Togal.AI building production LLM/RAG systems with sub-30ms inference.

United States

Message

At Togal.AI, I lead architecture and productionization for AI/ML systems supporting 20 production applications. I built LLM/RAG applications for document processing and knowledge retrieval, and helped reduce operational and processing time by approximately 25%.

I also lead distributed PyTorch training and engineer real-time model serving that achieves sub-30ms inference latency. Previously, at DataRobot, I developed streaming pipelines processing approximately 10M events per day; at Levatas, I built customer lifetime value and churn models across approximately 5M customer records.

Experience

Work history, roles, and key accomplishments

Current

Lead Machine Learning Engineer

Current

Togal.AI

Mar 2023 - Present (3 years 7 months)

Lead architecture, engineering, and productionization of ML and Generative AI systems supporting 20 production AI/ML applications. Architect and scale data platforms processing approximately 5TB/day, and design LLM/RAG applications for document processing and knowledge retrieval.

Machine Learning Engineer | MLOps & AI Systems

DataRobot

Oct 2019 - Feb 2023 (3 years 4 months)

Designed and productionized end-to-end machine learning systems across NLP, anomaly detection, predictive analytics, classification, and real-time decisioning. Architected Kafka streaming pipelines processing approximately 10M events/day and built low-latency model APIs using FastAPI, Redis, TorchServe, and ONNX Runtime.

Kafka Streaming Feature Engineering AWS Airflow GitLab CI CD Transformers BERT Zero Shot Classification Isolation Forest Autoencoder fastAPI Redis TorchServe ONNX Runtime Model Deployment Model Monitoring MLOps

Junior Data Scientist

Levatas

Aug 2016 - Sep 2019 (3 years 1 month)

Built customer lifetime value (CLTV) and churn prediction models across approximately 5M customer records, improving prediction performance by approximately 20%. Developed predictive models for forecasting, customer behavior, classification, and business analytics using Python and Scikit-learn.