$ whoami

Senior Machine Learning Engineer specializing in high-performance inference and agentic post-training. I optimize production-grade SOTA models for low latency and fine-tune autonomous agents for complex reasoning and tool use. Previously a history major who accidentally wandered into a terminal and never left.

$ cat /etc/experience

Adobe – Senior ML Engineer, Applied Research (Digital Imaging) 2025-present

• Architected inference engine for 20+ CV endpoints (Photoshop/Lightroom) serving 150+ RPS – masking, segmentation, image editing, 14B VLM, 1B DiT, VAE models
• Reduced P50 latency by 40% and cut annual deployment costs by $0.5M via PyTorch AOTI compilation, FP8 quantization
• Prototyped autonomous image processing agent with multi-step graph workflows and LLM-as-a-Judge evaluation on Langfuse
• pytorch tensorrt ray vllm triton k8s terraform

Block (Square) – Senior ML Engineer, Conversations AI 2022-2025

• Transformed rule-based Square Assistant into LLM-based agent with tool use, prompt engineering, and RAG
• Preference-aligned SFT using customer interaction signals as proxy reward – improved suggested reply acceptance by 30%
• Built agent evaluation suite on Langfuse benchmarking tool-use accuracy and reasoning trajectories
• Fine-tuned U2Net for product photography background removal (F1: 0.65 -> 0.90), 20x training speedup with DDP on 8 GPUs
• pytorch huggingface pytorch-lightning langfuse docker k8s

Visa – Senior Data Scientist, Payment Systems Risk 2019-2022

• Deployed TensorFlow risk model for ATM cashout fraud detection – $2M/yr in fraud protection savings
• Led cross-vertical DS community of practice; trained 70 engineers on internal data lake
• scala spark tensorflow kafka hive

$ cat /etc/education

Georgia Institute of Technology – M.S. Analytics 2019

• GPA: 4.0 – Deep Learning, ML, Bayesian Stats, Optimization, Computer Vision

Middlebury College – B.A. History & Mathematics 2015

• Middle Eastern History major, Math & Arabic minor
• Started coding after graduation, never stopped

$ cat /proc/skills

Core	Agentic AI, High-Performance Inference, Post-Training Data Strategy, LLM Eval
Methods	RAG, PPO/GRPO, SFT, Synthetic Data Generation
Domains	Autonomous Agents, VLMs, Diffusion, NLP
Stack	Python, PyTorch, Triton Server, Langfuse, Spark, Kafka, SQL, Docker, K8s

$ cat ~/stats.yml

• Marathon PB: 4:27:28
• Table Tennis: 2024 US Open Women’s Singles (30+) – 3rd Place
• Also into: biodynamic wine, photography, traveling

$ echo $CONTACT

GitHub – LinkedIn – dcvionwinnie@gmail.com