Building Production AI Systems with Python + GenAI | Measurable business impact from Day 1

Most entry-level candidates have tutorial projects. I have production code with measurable business impact.

Status: ✅ Live in production | 🌐 Public repository

Automated Python ETL pipeline for retirement plan distribution reconciliation at Daybright Financial.

Tech: Python • pandas • openpyxl • matplotlib • data validation • pytest • GitHub Actions CI • faker (synthetic data)

→ View Full Documentation & Code

Each project introduces new skills that build on the previous — from pure Python ETL to multimodal AI, RAG, statistical research systems, and an autonomous trading platform.

🏗️ Production Standard: Every project ships with architecture diagram (Mermaid), Dockerfile, evaluation metrics table, demo GIF, and "What I Learned" section.

🚩 Two flagships: Attention-Flow Catalyst (read-only small-cap swing research) and Crucible (autonomous intraday execution) — two genuinely different hard problems. Crucible is the chosen first build.

AI-Powered PII-Safe Data Intelligence | "Chat With Your Data"

Natural language analytics for retirement plan operations with PII protection and AI guardrails.

Tech: Python • pandas • Streamlit • Gemini SDK • PandasAI • Pydantic • DeepEval • Docker • GitHub Actions CI

AI-Powered HR Policy Chatbot | "Ask Your Policies" | 🔌 Exposes FastMCP server

RAG chatbot that answers employee policy questions with cited sources and auto-escalates to HR when uncertain.

Tech: Python • Anthropic SDK (primary, Gemini fallback) • ChromaDB • Gemini Embeddings • Streamlit • Pydantic • DeepEval • RAGAS • SelfCheckGPT • FastMCP • Docker • GitHub Actions CI

AI-Powered Distribution Form Validator | "From Paper to Processing"

Multimodal AI system that reads retirement plan distribution forms (handwritten checkboxes, signatures), validates against business rules, and routes results.

Tech: Python • Gemini Vision SDK • Streamlit • Pydantic • DeepEval • Docker • GitHub Actions CI

AI-Powered Workflow Demand Analysis | 🚧 In Development

Analyzing 8+ months of OnBase workflow data to enable data-driven staffing decisions with AI-powered natural language queries.

Tech: Python • pandas • Streamlit • Gemini SDK • PandasAI • Plotly • DeepEval • Docker • GitHub Actions CI

→ View Project

AI-Powered Streaming Subscription Rotation Advisor | "Spend Less, Watch More"

Consumer-facing dashboard that optimizes streaming subscriptions through AI-driven rotation scheduling, cost-per-view analytics, and content search.

Tech: Python • httpx async • Watchmode/TMDB APIs • Streamlit • Gemini SDK • Pydantic • DeepEval • LangSmith • Docker • GitHub Actions CI

AI-Powered Predictive Trigger Analysis for Small-Cap Stocks | 🚧 Phase 1A Active

Research Question: Which trigger or combination best predicts +10% price moves within 3 trading days?

Flagship project designed as a defensible research system with statistical methodology — evolves through all 5 career stages.

• ✅ Dynamic stock screener with survivorship bias controls
• ✅ 3+ years data collection for 50+ stocks
• ✅ Walk-forward validation (train Y1-2, test Y3)
• ✅ Bootstrap 95% confidence intervals
• ✅ Trigger leaderboard ranked by hit rate
• ✅ Forward signal generator

Tech: Python • DuckDB • Parquet • httpx async • edgartools • Anthropic SDK (primary, Gemini/OpenAI fallback) • PandasAI • Streamlit • DeepEval • SelfCheckGPT + FActScore • Docker • GitHub Actions CI

→ View Project

Autonomous Intraday Trading Research Platform | 🚧 Phase 1 | 🦙 Local-First AI

The question it answers, for any strategy: Does this have a real edge that survives out-of-sample validation — and can an autonomous agent trade it without me babysitting it?

Strategy-agnostic platform that takes any intraday strategy through three validation gates: backtest → paper → live. An LLM research analyst proposes strategy improvements, but its ideas are proved by deterministic backtests it never optimizes against — behind a sealed out-of-sample vault with every peek logged in an overfitting-budget ledger. That's what makes it defensible rather than an overfit black box.

Distinct from AFC (why two flagships, not redundancy): AFC is read-only research on illiquid sub-$5 small-caps over a multi-day swing horizon; Crucible is autonomous execution on liquid names over an intraday horizon. ~70% shared engineering spine, two different hard problems.

Tech: Python • own event-driven backtest harness → NautilusTrader (LGPL, free) • Optuna • DuckDB • Parquet • Ollama/Qwen3 (local-first) → Gemini → Anthropic → OpenAI • Pydantic • LangGraph • Alpaca (paper + live) + Schwab Trader API/TOS (live) • DeepEval • Docker • GitHub Actions CI

⚖️ Educational/research project. Not investment advice; makes no claim of positive expectancy — validation is the entire point.

→ View Project

📚 Data Portfolio Hub Central repository linking all projects with business context, technical details, and impact metrics.

Languages & Dev Tools

Data Analysis & Processing

Data Collection

Visualization & BI

Testing & CI/CD

Containerization

AI & GenAI Tools

AI Evaluation Frameworks

Trading & Backtesting (Crucible)

APIs & Domain Expertise

Domain Expertise: 15+ years business data • 4 years financial services • 6 years active trading

Planned skills for GenAI Data Engineer + AI Systems Architect role (Months 6-15):

Cloud:         AWS (S3, Redshift, Lambda, Glue), BigQuery
Containers:    Docker & Kubernetes Masterclass
Orchestration: Apache Airflow, dbt (data build tool)
Big Data:      PySpark, distributed computing
Databases:     PostgreSQL, Vector DBs (Pinecone/Weaviate/Qdrant)
AI Systems:    RAG infrastructure, embedding pipelines, unstructured data ETL
Certification: AWS Certified Data Engineer Associate

• 🚧 CS50: Introduction to Computer Science (Harvard)
• 🚧 Python for Everybody Specialization (University of Michigan)
• 🚧 Google Data Analytics Professional Certificate
• 🚧 IBM Data Analyst Professional Certificate (11 courses)
• 📅 Statistics with Python Specialization (University of Michigan)

• 🚧 LLM SDK Engineering — Provider-agnostic architecture (Gemini, OpenAI, Claude)
• 🚧 IBM Generative AI Engineering Professional Certificate (16 courses) — RAG, LangChain, fine-tuning (LoRA/QLoRA), deployment (Stage 1 primary)
• 🚧 Generative AI Data Analyst Specialization (Vanderbilt) — ChatGPT+SQL workflows, CLUE/TRUST/CAPTURE frameworks
• 🚧 ChatGPT Prompt Engineering for Developers (DeepLearning.AI) — API basics, prompt design, few-shot learning
• 🚧 AI Python for Beginners (DeepLearning.AI) — Andrew Ng's AI-first Python foundation
• 📅 30 Days of Streamlit Challenge — Build AI UIs fast

• 📅 Building & Evaluating Advanced RAG (DeepLearning.AI) — RAG Triad metrics, evaluation-driven development
• 📅 Docker for Beginners with Hands-on Labs (KodeKloud/Coursera) — Dockerfile, Docker Compose, containerization

• ⚡ Cursor AI IDE — Primary editor with Composer mode
• ⚡ ChatGPT Plus — Advanced Data Analysis workflows
• ⚡ VS Code — Jupyter notebooks & data exploration

• 📈 6+ years active trading (swing & day strategies)
• 🌅 4:30 AM club (early morning focused study)
• ♟️ Chess enthusiast (strategy translates to markets!)
• 🤖 Fascinated by LLMs transforming financial analysis
• 📚 Reading: Machine Learning for Algorithmic Trading + Hands-On LLMs
• 🎯 Obsessed with data-driven decision making
• 💪 Proving systematic learning beats raw talent

📈 Stats & Contributions

💻 Most Used Languages

I'm Open To:

• 💼 Data Analyst & AI Engineer Opportunities (Remote preferred)
• 🤝 Code reviews and technical discussions
• 🎓 Knowledge exchange on data + AI + finance

Let's Connect If You:

• Value production code over tutorial completions
• Are hiring Data Analysts with proven delivery capability
• Are building data-driven AI systems