AI/ML Engineer

AI/ML Engineer — Synapse AI Solutions

San Francisco, CA | August 2023 – Present

• Trained and deployed a custom Llama-based LLM for customer support automation, reducing customer ticket resolution time by 32% while maintaining a 94% accuracy score.
• Optimized model inference using Triton Inference Server and TensorRT, reducing P99 latency by 45% and increasing throughput to 1,200 requests per second.
• Architected end-to-end MLOps pipelines using Kubeflow and Argo Workflows, cutting model retraining time from 4 days to 14 hours.
• Implemented real-time drift detection and automated data validation using Great Expectations, reducing silent model failures by 68% over 9 months.

Machine Learning Engineer — Helix Analytics Group

Denver, CO | June 2021 – July 2023

• Developed and shipped a deep learning recommendation engine using PyTorch, driving a 14% increase in user click-through rates (CTR) and a 9% rise in purchase conversions.
• Built feature store infrastructure using Feast, reducing model feature engineering latency from 120ms to 8ms and ensuring training-serving consistency.
• Managed a cluster of 32 AWS EC2 GPU instances, implementing spot instance scheduling to decrease monthly cloud compute spend by 28%.
• Partnered with 4 backend engineers to integrate predictive scoring endpoints into the core API, managing over 8 million daily scoring events.

Machine Learning, Deep Learning, Generative AI, Natural Language Processing (NLP), Computer Vision, PyTorch, TensorFlow, Python, Kubernetes, Docker, Triton Inference Server, TensorRT, Feast, Kubeflow, Argo Workflows, AWS (EC2, S3, SageMaker), GCP (Vertex AI), Git, CI/CD, SQL

Distributed LLM Finetuning Framework

Role: Lead ML Infrastructure Engineer

Tools: PyTorch, DeepSpeed, Ray, Kubernetes, AWS

Built a training pipeline that fine-tuned 7B-parameter models on custom enterprise datasets, reducing training time from 72 hours to 18 hours using DeepSpeed ZeRO-3.

Real-Time Object Detection Pipeline

Role: ML Engineer

Tools: TensorFlow, OpenCV, TensorRT, Triton, Docker

Developed a pipeline processing 30fps video streams, achieving model optimization that reduced inferencing GPU memory consumption by 34%.

• Google Professional Machine Learning Engineer (2024)
• AWS Certified Machine Learning – Specialty (2022)

• Programming Languages: Python, C++, Go, SQL
• Tools & Frameworks: PyTorch, TensorFlow, Triton, Docker, Kubernetes, Feast, Kubeflow, Ray, Great Expectations
• Availability: 2 weeks notice