Data & ML Engineer building production-grade pipelines, intelligent agents, and ML systems. MCP · LangGraph · RAG · LLMs.
I work at the intersection of data engineering, machine learning, and LLM systems — turning raw data into reliable, scalable, and intelligent products. My work spans ETL infrastructure processing terabytes daily, production fraud detection models, fintech analytics, and AI agents powered by large language models.
- Data Engineering — Apache Airflow, Delta Lake, Great Expectations, SageMaker Feature Store
- Machine Learning — XGBoost, SMOTE, MLflow, SHAP, A/B testing, FastAPI model serving
- Analytics — SQL window functions/CTEs, KPI automation, fintech metrics at scale
- LLM / Agents — LangGraph ReAct agents, LoRA/QLoRA fine-tuning, RAG pipelines, Claude API, MCP servers
| Project | Description | Stack |
|---|---|---|
| blockchain-etl-pipeline | 18 Airflow DAGs, 3.2TB/day, 99.4% SLA — Delta Lake time-travel, Great Expectations, SageMaker Feature Store | Python · Airflow · Delta Lake |
| fraud-detection-ml | 6-model XGBoost benchmark, SMOTE, MLflow, SHAP — 85% precision on 0.3% fraud rate | Python · XGBoost · FastAPI |
| claims-forecasting-platform | End-to-end claims demand forecasting & staffing optimization — SARIMA, XGBoost, anomaly detection, automated capacity reports | Python · statsmodels · XGBoost · SQLite |
| fintech-kpi-analytics | A/B testing (n=12K), SQL CTEs, automated KPI reporting — cut ad-hoc requests 38% | Python · SQL |
| langgraph-analyst-agent | Multi-step ReAct agent with LangGraph: 5 tools, automated reports — turnaround 3 days → 4 hours | Python · LangGraph |
| llm-finetuning-pipeline | End-to-end LLM fine-tuning with LoRA/QLoRA, W&B tracking, ROUGE evaluation | Python · HuggingFace |
| iedr-utility-data | Scalable data lakehouse ingesting electrical-circuit & DER data from NY utilities — Bronze/Silver/Platinum Delta layers, FastAPI endpoints | Python · Databricks · PySpark · Delta Lake |
| scisynth | AI research synthesis — arXiv search, Claude-powered analysis, RAG chat, literature review generation | TypeScript · Claude API |
| mcp-finance-server | MCP server exposing financial intelligence tools to Claude — company lookup, AML risk scoring, sanctions screening (OFAC/UN/EU/UK), document intelligence | TypeScript · MCP · Node.js |
| agentic-kyb-pipeline | Multi-agent KYB/AML pipeline: entity resolution → sanctions screening → adverse media → risk report with full audit trail | Python · LangGraph · Pydantic |
- Email: avinashpolina2028@gmail.com
- GitHub: avinashp80089-del