📍 Chicago, IL ·

"Most AI demos break. I build the systems that don't."

I spent 13 plus years building IT infrastructure before I moved into AI. That background shapes how I work: I don't just build things that demo well. I build things that hold up under edge cases, under load, and over time.

Right now I'm focused on LLM systems, RAG pipelines, and the evaluation infrastructure that makes AI actually trustworthy in production.

LLM Systems & Prompt Engineering Few-shot, Chain-of-Thought, and ReAct prompting. RAG pipelines with retrieval, grounding, and structured output. LLM evaluation frameworks, hallucination detection, and benchmarking across GPT-4, Claude, LLaMA, and Gemini.

AI Automation & Systems Design End-to-end workflow automation in Python. Multi-step AI pipelines designed for real-world failure modes. Human-in-the-loop (HITL) architectures. Prompt chaining with structured reasoning.

Data Engineering & Analytics SQL analytics systems in PostgreSQL and SQL Server. ETL pipelines, star schema modeling, KPI design. Cohort analysis, retention modeling, Power BI dashboards with DAX.

Machine Learning Classification, regression, and clustering with scikit-learn. Feature engineering, model evaluation, fraud detection, and churn prediction pipelines.

A production-grade harness for testing LLM prompts systematically ,not by eyeballing output.

• Four-dimensional weighted scoring (JSON structure, required keys, safety compliance, content rules).
• Simulated OpenAI provider for continuous testing without API costs. 100% pass rate on test suite.
• Reduced hallucinations 30–40% through structured prompt optimization.
• Tech: Python OpenAI API Prompt Engineering LLM Evaluation

Built an AI-powered Prompt Engineering Evaluation System that analyzes, scores, compares, and refines LLM prompts using structured evaluation metrics instead of manual guesswork.

• Designed a multi-dimensional prompt scoring framework measuring clarity, specificity, robustness, efficiency, and alignment, including hallucination-risk analysis and AI-generated optimization suggestions.
• Implemented prompt versioning and head-to-head comparison workflows to track iterative improvements and evaluate prompt performance across different task types (RAG, summarization, reasoning, extraction, etc.). -Developed a full-stack React application integrated with the Anthropic Claude API for real-time prompt evaluation, refinement, and predicted output simulation.
• Tech: React18 JavaScript Prompt Engineering LLM workflows Tailwind CSS Anthropic Claude API (Claude Sonnet 4) AI Evaluation Systems

End-to-end analytics pipeline modeling user performance & engagement

• Leaderboard, streak & rolling 7-day metrics
• Question difficulty modeling & query optimization
• Tech: PostgreSQL Advanced SQL Window Functions KPI Modeling

E-Commerce Analytics Platform (SQL-Only) Complete SQL-based analytics system for transactional data

• Revenue, AOV, LTV & retention metrics
• Cohort analysis & reusable reporting views
• Tech: PostgreSQL SQL Aggregations Window Functions

Customer Churn Prediction End-to-end ML pipeline for churn prediction using telecom data

• Feature engineering, model training & evaluation
• Business-driven retention insights
• Tech: Python Pandas scikit-learn Classification

Credit Card Fraud Detection Fraud detection model addressing severe class imbalance

• Precision/Recall optimization & F1-score evaluation
• False positive minimization strategy
• Tech: Python Pandas scikit-learn

NLP Chatbot (Flask App) Rule-based chatbot with TF-IDF & cosine similarity

• REST API backend & preprocessing pipeline
• Tech: Python Flask NLTK HTML CSS

• Production-grade RAG systems with evaluation scoring
• LLM workflow automation pipelines.
• ML monitoring and drift detection.
• Advanced prompt optimization frameworks.

Full-time remote roles in Generative AI Engineering, Prompt Engineering, AI Systems Engineering, LLM Application Development, and Data Engineering.

Portfolio · LinkedIn · Email