tyler cooper

N

MLOps Engineer

Neudesic

Aug 2023 - Current (3 years)

Designed and deployed a Kubernetes-based private ML cloud supporting distributed model training and inference pipelines. Implemented GPU resource discovery and scheduling to improve utilization across multi-tenant workloads. Automated policy enforcement and compliance using Kyverno and admission webhooks. Managed Kubeflow Pipelines for training, hyperparameter tuning, and deployment via Katib, Helm, and ArgoCD. Integrated Prometheus + Grafana + ELK for unified observability and performance tracking. Established GitOps-based CI/CD for ML workflows, ensuring reproducible deployments and scalable infrastructure.

AI/ML Engineer

Hugging Face

Mar 2020 - Aug 2023 (3 years 6 months)

Fine-tuned Qwen-2.5, GPT-3.5-Turbo, and Phi-3-Vision models for multilingual conversational, financial, and document-understanding tasks using LoRA and DPO. Built retrieval-augmented NLP pipelines using FAISS + LangChain, enabling low-latency document Q&A and RAG workflows. Developed real-time forecasting pipelines combining Prophet, LSTM, and Transformers, deployed via Triton Inference Server. Created data curation and annotation tools using PostgreSQL and PySpark, improving fine-tuning efficiency. Optimized LLM throughput and latency using vLLM and concurrent inference strategies.

Machine Learning Engineer

FLEXERA

Jul 2019 - Mar 2020 (9 months)

Built video content moderation systems using temporal action recognition and multi-label classification, boosting accuracy by 18%. Developed object detection and segmentation pipelines for brand-safety filtering; automated retraining using Kubeflow and active learning. Leveraged CLIP vision-language embeddings for dataset clustering and weak supervision, reducing annotation effort. Built model-assisted labeling pipelines, reducing manual labeling time by over 32%.

ML Engineer

iNexxus

Nov 2017 - Jul 2019 (1 year 9 months)

Developed OCR and ANPR systems with >90% accuracy using PyTorch + OpenCV. Deployed real-time number plate recognition from RTSP streams with optimized frame batching. Built TorchScript-based car damage detection models for Android, optimized for latency and precision. Applied Pix2Pix GANs for visual enhancement and shadow removal, improving detection robustness in low-light environments.

MLOps Engineer

Palantir

Jan 2015 - Jan 2017 (2 years 1 month)

Designed scalable ML pipelines using Kubeflow and Airflow for edge and cloud inference systems. Deployed deep learning models for camera-edge devices using TensorRT and Triton, enabling low-latency inference. Built CI/CD pipelines for retraining and deployment with GitHub Actions, Docker, and Kubernetes. Integrated Prometheus + Grafana dashboards to monitor drift, latency, and data quality. Developed hybrid pipelines combining financial forecasting, NLP, and edge deployment for experimental analytics platforms.

Machine Learning Engineer (Finance & Edge AI)

QuantConnect

Jan 2014 - Jan 2015 (1 year 1 month)

Developed multi-horizon stock forecasting systems using Prophet, LSTM, GRU, and Transformer architectures, integrating FinBERT-based news sentiment as exogenous input. Designed financial NLP pipelines for sentiment and risk analysis using FinBERT and Hugging Face Transformers. Implemented automated retraining workflows with MLflow + DVC + Airflow + Optuna, improving maintainability. Deployed real-time APIs via BentoML for prediction serving on cloud and edge environments. Integrated Evidently AI for model drift detection and Feast for feature versioning and monitoring.

About

Detail

Timeline

Résumé