Ron Christopher White

SENIOR MACHINE LEARNING ENGINEER

Replicant

Jan 2023 - Current (3 years 6 months)

• Spearheaded the design and deployment of LLM-driven Conversational AI systems using GPT-3.5, Claude, and Llama 2, powering 24/7 customer support at scale across telecom and healthcare sectors. • Built robust RAG (Retrieval-Augmented Generation) pipelines using LangChain, FAISS, and Pinecone, enhancing response grounding and cutting hallucinations by 40%. • Developed fine-tuned LLM variants via PEFT (LoRA, QLoRA) and quantization techniques (4-bit, 8-bit) for task-specific inference with sub-500ms latency. • Orchestrated scalable LLM serving infrastructure using Triton Inference Server, ONNX, and multi cloud GPU clusters (AWS/GCP), achieving a 60% latency improvement under real-world load. • Led the implementation of end-to-end LLM ML

SENIOR AI SOFTWARE ENGINEER

IBM

Jun 2021 - Dec 2022 (1 year 7 months)

• Designed a scalable Document Intelligence System processing 50K+ pages daily using a hybrid approach: ◦ Computer Vision: Implemented Faster R-CNN (TensorFlow) for layout detection (98% recall). ◦ NLP: Developed custom BERT variants (Hugging Face) for entity extraction (93% F1- score). ◦ Integrated SpaCy dependency parsing for rule-based post-processing. • Built a Time-Series Forecasting Platform for supply chain optimization: ◦ Integrated ensemble models: Prophet, ARIMA, and LSTM. ◦ Automated feature engineering using tsfresh. ◦ Reduced forecast error by 22% compared to legacy systems. • Developed a comprehensive Model Governance Framework: ◦ Tracked Data Drift using KS tests. ◦ Monitored Concept Drift via performance m

DATA SCIENTIST

Blue Yonder

Feb 2017 - Apr 2021 (4 years 3 months)

• Created a robust Demand Forecasting Engine processing: ◦ 10TB+ weekly POS, inventory, and promotion data. ◦ Utilized XGBoost/LightGBM models with automated hyperparameter tuning via Optuna. ◦ Achieved 90%+ accuracy across 200+ SKUs in 15 retail chains. • Built an Anomaly Detection System for logistics: ◦ Combined Isolation Forest with GAN-based approaches. ◦ Reduced false positives by 30% compared to threshold-based methods. • Designed Spark optimizations on AWS EMR clusters: ◦ Implemented Parquet partitioning strategies. ◦ Optimized Broadcast Joins. ◦ Cut processing costs by 40%.

JUNIOR SOFTWARE ENGINEER

Cisco

Sep 2015 - Dec 2016 (1 year 4 months)

• Created a Real-Time Dashboard using Flask + D3.js for support analytics.

Ron Christopher White

About

Detail

Timeline

Résumé