AI & Automation Consultant | Data Scientist | ML Engineer
Houston, TX
I architect production AI systems that turn complex data problems into measurable business outcomes - spanning ML pipelines, LLM-powered automation, and end-to-end system design.
Currently at Bridgital - building distributed automation systems, generative AI agents with compliance guardrails, and semantic search pipelines on AWS Bedrock.
Previously - OCR and document classification systems at Bitpoint (75% reduction in manual entry across 500+ daily medical bills) and a multi-domain legal RAG system at Hartsfield & Associates (91% Recall@5 across 1,500+ documents).
Education - MS in Data Science (4.0 GPA) | BE in Mechanical Engineering | Peer-reviewed publication on queuing theory
| Domain | Technologies |
|---|---|
| Languages | Python, SQL, JavaScript, TypeScript, R |
| ML & AI | PyTorch, TensorFlow, Scikit-learn, HuggingFace, LangChain, LangGraph, CrewAI |
| LLMs & GenAI | OpenAI, Claude, Groq, AWS Bedrock, RAG architectures, Fine-tuning |
| Cloud & Infrastructure | AWS (Bedrock, S3, RDS, App Runner, Amplify), Docker, PostgreSQL, MongoDB, Redis, FAISS |
| Automation | n8n, Zapier, Power Automate |
| Visualization | Tableau, Power BI, Plotly, D3.js |
| Project | Impact | Stack |
|---|---|---|
| Enterprise Field Services Automation | 80% reduction in administrative overhead | n8n, Zapier, Monday.com, Gemini AI |
| Legal Research AI | 91% Recall@5, sub-3s query latency across 1,500+ docs | Groq LLaMA 3.1, FAISS, Sentence-Transformers |
| LoanPilot Social | RESPA & NMLS compliant AI content generation | OpenAI, LangChain, Next.js 14 |
| TaxIntake | AES-256 encrypted full-stack client onboarding | React, Node.js, PostgreSQL, AWS, n8n |
| PodBrain | Semantic attendee matching via 27-point logic matrix | AWS Bedrock, Amazon Titan, FastAPI |
| CyberOutreach | B2B multi-agent GTM with persona-specific campaigns | Claude 3.7 Sonnet, Apollo.io, HubSpot |
| Project | Result | |
|---|---|---|
| Hybrid Residual LSTM Forecasting | Gold price prediction - 0.37% MAPE, bidirectional LSTM + Attention | Repo |
| Cognitive Document Retrieval | RAG pipeline with vector-semantic search and Groq inference | Repo |
| Autonomous Financial Multi-Agent | LLaMA 3.2-based multi-agent system for market research | Repo |
| Time Series Benchmarking | ARIMA, SARIMA, SARIMAX, and Prophet comparison study | Repo |
Contributions merged into llama.cpp, Optuna, River, QuTiP, and Haystack, with PRs under review at PyTorch, Keras, LlamaFactory, Lightning, AutoGluon, and ChromaDB.
Merged
| Project | Contribution | PR |
|---|---|---|
ggml-org/llama.cpp |
Fix WebUI CORS proxy missing API key for MCP connections | #21193 |
optuna/optuna |
Move type-only imports into TYPE_CHECKING block |
#6609 |
online-ml/river |
Per-sample weights in progressive validation | #1795 |
qutip/qutip-qip |
Default zero state for QubitCircuit.run() |
#376 |
deepset-ai/haystack |
Update llama.cpp repository links to ggml-org | #10964 |
Under Review
| Project | Contribution | PR |
|---|---|---|
pytorch/pytorch |
Fix misleading out_dtype signatures in bmm/mm/addmm docs |
#179182 |
keras-team/keras |
Fix DepthwiseConv1D channels_first dilation bug |
#22641 |
hiyouga/LlamaFactory |
Fix json.load failure with PosixPath |
#10361 |
Lightning-AI/lightning |
Fix CometLogger offline mode bug | #21634 |
autogluon/autogluon |
Custom cv_splitter for TabularPredictor |
#5628 |
chroma-core/chroma |
Azure OpenAI embedding deployment tests | #6760 |
- AWS Certified AI Practitioner - Amazon Web Services (Verify)
- MLOps: Getting Started - Google Cloud
- Transformer Models and BERT - Google Cloud
- Encoder-Decoder Architecture - Google Cloud
- AI Demystified - IBM WatsonX
I write a weekly research publication breaking down the latest AI papers and automation frameworks into actionable insights for practitioners.