class MLEngineer:
def __init__(self):
self.name = "Ahmeduddin Mohammed"
self.role = "ML / AI Engineer"
self.location = "Harrison, New Jersey"
self.education = "MS Computer Science — NJIT"
self.specialties = [
"Production ML Pipelines",
"Causal Inference Engines",
"LLM Agent Orchestration",
"Deep Learning Systems",
]
self.stack = {
"ML": ["PyTorch", "LSTM", "Transformer", "XGBoost", "SHAP"],
"Causal": ["DoWhy", "EconML", "CausalForestDML", "PSM", "DiD"],
"LLM": ["LangGraph", "LangChain", "GPT-4o-mini", "RAG", "ChromaDB"],
"MLOps": ["MLflow", "Docker", "GCP Cloud Run", "GitHub Actions"],
"Backend": ["FastAPI", "PostgreSQL", "Redis", "Pydantic v2"],
}
def pitch(self):
return """
ML Engineer specializing in production systems — causal inference,
deep learning, and LLM orchestration. Every project ships with
full test coverage, real metrics, and a live deployment.
"""
| 🎯 Best AUC | ⚡ Inference | 📊 Dataset | 🤖 Pipeline | ✅ Tests | 📦 Deployments |
|---|---|---|---|---|---|
| 0.9868 | 27ms p95 | 14M rows | 0.36s | 369 | 4 Live |
| SessionScout LSTM | Redis cached | LiftLab Criteo | CareAgent GCP | 4 ML projects | HuggingFace + GCP |
4-model cascade (LR → XGBoost → LSTM → Transformer) predicting mid-session purchase probability at sub-30ms latency. Confidence-gated escalation cuts GPU inference cost by 60%. Bidirectional LSTM (256 hidden, 2-layer) + 4-head self-attention over 121 engineered features.
PyTorch LSTM Transformer XGBoost FastAPI Redis Docker HuggingFace
5 parallel causal estimators (PSM, DiD, T-Learner, X-Learner, CausalForestDML) on 14M-row Criteo dataset. Per-customer CATE with confidence intervals. CausalForestDML honest splitting prevents bias. MMD drift test flagged distribution shift at 0.136.
DoWhy EconML CausalForestDML SHAP MLflow PSI MMD Streamlit
XGBoost scorer + SHAP top-5 risk factors mapped to CFPB Reg-B codes via ChromaDB RAG (9,977 regulatory chunks). GPT-4o-mini generates legally-grounded adverse action notices at temperature=0 for legal determinism. Cosine similarity threshold 0.7 blocks hallucination.
XGBoost SHAP LangChain ChromaDB GPT-4o-mini dbt PostgreSQL MLflow
LangGraph StateGraph routing 5 specialized agents over 10,000 CMS Medicare providers. Supervisor → DataCleaner → Statistical → Isolation Forest Anomaly → Summarizer → Reporter. Pre-reserved schema fields prevent inter-agent column bugs. Template fallback guarantees pipeline completion.
LangGraph LangChain GPT-4o-mini Isolation Forest GCP Cloud Run PostgreSQL MLflow
| Project | Description | Stack | Links |
|---|---|---|---|
| AI Financial Analyst | RAG variance engine — deterministic decomposition + GPT citation grounding · 100% reconciled | FastAPI · React · OpenAI · pdfplumber · Vercel |   |
| AI Money Brain | Multi-bank finance copilot — 5+ banks, GPT-4 categorization, 192 tests, 92% coverage | FastAPI · PostgreSQL · GPT-4 · JWT · Docker · Alembic |  |
| Portfolio Tracker | Full-stack SaaS dashboard — auth, Recharts analytics, dark/light mode, Vercel CI/CD | React · TypeScript · Recharts · Vite · Vercel |  |
ML & Deep Learning
Causal Inference
LLM & Agents
MLOps & Cloud
Backend & Data
