Derek S.

MLOps Engineer

Plaid

Jan 2026 - Current (10 months)

- Built Graph Neural Network (GNN) fraud detection models using Python, PyTorch Geometric, and AWS SageMaker to identify complex transaction network risks and reduce fraud losses.
- Developed real-time feature pipelines with Apache Flink, Kafka, Feast, and Redis, enabling low-latency fraud detection at scale.
- Scaled ML inference infrastructure using Triton Inference Server, C++, AWS EKS, and Kubernetes for high-volume transaction processing.
- Applied transformer-based NLP models with Hugging Face to normalize merchant data and improve fraud graph accuracy.
- Created automated model evaluation and hard-negative mining frameworks using PyTorch and Scikit-learn to continuously improve fraud detection performance and comp

Senior Software Engineer

Plaid

Aug 2025 - Dec 2025 (5 months)

- Built a bank data enrichment system using Go, Python, and FastAPI with optimized Goroutine concurrency.
- Cut AI fees by using Redis Enterprise and Milvus Vector Database to cache OpenAI answers.
- Sped up GPU analysis via TensorRT INT8 quantization on NVIDIA Triton Inference Server.
- Protected client apps from AI outages using Go circuit breakers and RE2 regex matching.
- Improved spike responsiveness by delegating validation and tokenization to Celery and RabbitMQ.
- Blocked millions in fraud using Go, confluent-kafka-go, and Apache Kafka on AWS MSK.
- Reduced network bandwidth costs by migrating from REST/JSON to gRPC and Protobuf v3.
- Accelerated risk assessments by running XGBoost and LightGBM via

Artificial Intelligence Engineer

turing

Mar 2025 - Jul 2025 (5 months)

- Engineered Python and SQL training pipelines to clean complex architectural datasets for AWS and PostgreSQL.
- Debugged hundreds of C++ training prompts to ensure proper STL container usage and C++20 standards.
- Implemented a Go, Docker, and Bash code-verification sandbox to validate and benchmark LLM outputs.
- Executed RLHF workflows in Labelbox to rank model responses on Go concurrency safety and PEP-8.
- Developed Bash and Python utility scripts using Labelbox APIs, increasing data throughput by 50%.
- Designed Chain-of-Thought prompting strategies for AWS CloudFormation and Terraform infrastructure logic.
- Used SQL gap analysis on datasets to optimize model reasoning across complex distributed systems.

Founding Engineer

EPAM Systems

Aug 2024 - Mar 2025 (8 months)

- Built a full-stack sales intelligence MVP using Next.js, TypeScript, FastAPI, and Python.
- Automated background profile discovery and converted raw contact data into personalized event ideas.
- Architected an asynchronous data enrichment and event generation pipeline using Celery and Redis.
- Engineered a multi-step FastAPI and Pydantic ingestion system for CSV uploads and validation.
- Implemented a data discovery workflow using SerpAPI and OpenAI API to extract structured signals.
- Enforced data reliability using OpenAI Structured Outputs with Pydantic for confidence scoring.
- Designed a custom PostgreSQL scoring algorithm to rank recommendations into priority tiers.
- Integrated generative AI promp

Software Engineer

CVS Health

Dec 2017 - Jul 2024 (6 years 8 months)

- Built a high-concurrency, HIPAA-compliant vaccine portal using React, TypeScript, Node.js, and Express.
- Created smooth step-by-step patient forms that scaled effortlessly during massive traffic surges.
- Prevented appointment double-booking errors by implementing distributed Redis atomic slot-locking.
- Accelerated patient checkout speeds by optimizing state transitions with Redux Toolkit.
- Introduced strict request validation rules using Zod to boost form submission performance.
- Engineered a real-time mail-order prescription tracking dashboard using Python and Django.
- Decoupled legacy mainframe lookups by building a Celery and Redis asynchronous pipeline.
- Standardized responsive user interfaces

About

Detail

Timeline

Résumé

Derek S.

About

Detail

Behavioral traits