Atlanta senior data scientist & ML engineer. I architect multi-agent orchestration, RAG pipelines, fraud ML, and production MLOps — and lead PrismBase.ai, a data science and AI consulting company.
Founder
PrismBase.ai
Focus
ML · AI · RAG
9
Years in Data
6
Years in DS/ML/AI
7
Production AI Systems
01 — About
Nine years in data, six in DS/ML/AI — from AutoML frameworks to multi-agent orchestration and enterprise RAG.
As a data scientist proficient in Python, SQL, and machine learning, I specialize in developing end-to-end AI solutions that solve real-world problems. My work spans from traditional statistical modeling to cutting-edge deep learning and computer vision.
I'm passionate about building scalable ML systems — from AutoML frameworks to retrieval-augmented generation pipelines. Every project is designed for production, not just proof-of-concept.
I'm also the founder of PrismBase.ai, helping organizations turn their data into intelligent, production-ready solutions.
Deep learning, neural networks, and advanced machine learning at production scale.
Large-scale pipelines, ETL architecture, and real-time data infrastructure.
Translating complex analytics into strategic decisions that move the needle.
Leading an AI consulting company delivering production-ready solutions.
02 — Portfolio
Production-grade AI systems — multi-agent orchestration, RAG, fraud scoring, and enterprise GenAI. Each includes a deep-dive article.
01Multi-AgentMulti-agent inquiry automation for inbound business traffic. LangGraph orchestration, local Ollama inference, PostgreSQL audit trails, Redis hot state, and a Next.js control center.
02Enterprise GenAIEnterprise GenAI workbench with multi-agent routing, hybrid RAG, local LLM support, fine-tuning pipelines, and MLOps tooling — Next.js frontend with client-side demo orchestration.
03Local-First RAGLocal-first RAG stack with dual Chroma collections (public encyclopedia + papers), FastAPI retrieval pipeline, Ollama embeddings, citation-grounded answers, and Next.js query UI.
04Fraud ML05Google ADKRunnable Google ADK agent app with multi-agent transfer, RevOps lead triage, BSA/AML alert orchestration, tool-grounded résumé facts, and a React trace-replay demo UI.
06Policy-as-CodeMulti-agent fraud triage with OPA policy-as-code, Temporal workflows, hash-chained audit trails, HMAC-signed evidence, FastAPI RBAC, and a React operator console.
07Real-Time MLSports betting analytics platform with BERT sentiment analysis, RFM customer segmentation, churn prediction, real-time WebSocket feeds, and a FastAPI + Next.js stack.
Comprehensive portfolio showcasing various data science projects and methodologies. Features interactive Jupyter notebooks demonstrating machine learning techniques and data analysis workflows.
Local Retrieval-Augmented Generation (RAG) pipeline using FAISS for vector search and GPT-2 model. Demonstrates end-to-end RAG implementation without external APIs.
ZazuML fork - Easy AutoML solution for Object Detection. Streamlines the machine learning workflow from data preprocessing to model deployment.
Executive-quality reporting dashboard using fake generated numpy data. Demonstrates Streamlit's capabilities for business intelligence and data visualization.
End-to-end AI development project focused on computer vision applications. VisionDetect demonstrates practical implementation of AI in visual recognition tasks.
Collection of generative AI experiments and implementations. Explores various approaches to AI-generated content and creative applications.
Deep Dives
Architecture, tech stack, and setup — optimized for search and technical review.
Browse All ArticlesWriting
Data science, machine learning, and applied AI. For site-native posts and project stories, visit draketalley.ai/blog.
Read on Medium03 — Expertise
Full-stack data science — from statistical modeling to production ML infrastructure.
Discovery calls for consulting, architecture reviews, and production ML scoping — or send a brief and I'll reply within one business day.
Schedule
30-minute intro calls for consulting and architecture reviews.