Hi, I'm Shelton Simbi GenAI Engineer · LLM Systems Architect · Data Scientist · Based in New York
Most AI systems look great in a notebook. I build the ones that hold up in production.
I specialize in agentic AI, RAG pipelines, and LLM engineering, designing systems that combine large language models, vector retrieval, and multi-agent orchestration into applications that deliver real business value. I also teach these systems at graduate level at Yeshiva University, where I help engineers go from LLM basics to production deployments.
What I build:
- Agentic AI systems using LangGraph, CrewAI, and MCP (Model Context Protocol)
- RAG pipelines (hybrid, corrective, and graph-based) that scale beyond demos
- LLM powered applications evaluated with LangSmith and Ragas
- End-to-end ML systems from training through monitored production APIs
Core Stack
GenAI & Agentic AI
LLMs, RAG, Graph RAG, Hybrid RAG, Agentic AI, Multi Agent Systems, LangChain, LangGraph, LlamaIndex, CrewAI, MCP, Prompt Engineering, Context Engineering, Fine-Tuning, Embeddings, Semantic Retrieval
Evaluation & Safety
LangSmith, Ragas, DeepEval, Guardrails AI
Vector Databases
Pinecone, Weaviate, pgvector, Neo4j
Machine Learning
Scikit-learn, TensorFlow, PyTorch, Anomaly Detection, Predictive Modeling, NLP, Time Series Forecasting
Languages
Python, TypeScript, SQL
Backend & MLOps
FastAPI, Flask, Docker, vLLM, MLflow, REST APIs, CI/CD
Cloud & Infrastructure
AWS (Bedrock, SageMaker, EC2, S3), Azure, Vercel
Databases
PostgreSQL, MySQL, MongoDB, Snowflake, Apache Spark, Microsoft SQL Server
Frontend
Next.js, React, Streamlit
Selected Projects
ExamPrep — AI Exam Preparation Platform Turn course materials into graded practice. Ingests PDF, DOCX, and TXT files, generates MCQ, short answer, and scenario questions with semantic deduplication, runs timed exams, and delivers rubric-aware feedback. Stack: LLMs, FastAPI, Next.js, PostgreSQL, Docker
Nash Paint — AI Powered Paint Consultation System RAG powered assistant delivering intelligent product recommendations, image-based color extraction, virtual repaint visualization, branch lookup, and customer support. Stack: LLMs, RAG, Embeddings, FastAPI, Computer Vision
NSSA AI Powered Business Intelligence Platform Natural language to SQL querying over enterprise pension fund databases with LLM-based reasoning, automated reporting, and intelligent dashboards. Stack: LangChain, RAG, FastAPI, PostgreSQL, Streamlit
AI Legal Analytics — Law Capstone Predicts appeal outcomes using a model trained on 18,500+ cases. Hybrid retrieval with LegalBERT and BM25 rank fusion surfaces relevant precedents. Generates structured appellate briefs grounded in winning cases. Stack: LegalBERT, MLP Classifier, FastAPI, Next.js 14
Achievements
- 47% improvement in fraud detection accuracy at NSSA
- 60% reduction in data pipeline processing time
- 35% reduction in operational forecasting errors at JSC
- Neo4j Fundamentals Certified — April 2026
Connect
- Portfolio: https://personal-portfolio-steel-phi-27.vercel.app/
- LinkedIn: https://www.linkedin.com/in/shelton-simbi-750155204/
- Email: ssimbi@mail.yu.edu