AI/ML Engineer · Data Scientist · Data Analyst — Python · LLMs (LoRA/QLoRA) · NLP · MLOps · SQL · GIS

• AI Engineer Intern @ Global Alliant — Deployed OCR → JSON pipeline (LayoutLMv3 + Mistral-7B/DollyV2) that processed 2,300+ invoices end-to-end; added eval harness & sampling audits; reduced fine-tuning cost ~98% with LoRA/QLoRA (on T4), improving throughput and reliability for downstream finance ops.

• Research Assistant, University of Maryland — Built forecasting + explainability for socio-economic & geospatial data (ARIMA/Prophet/XGBoost + SHAP), delivering −35% MAPE (18.6 → 12.1) and −22% RMSE vs baselines with reproducible, parameterized notebooks.

• Parcel-Equity (Lubbock) — 300k-parcel analytics (1945–2021) — Designed ETL/QA to 99% data quality; OLS/GLM + K-Means + ArcGIS Pro visualizations; quantified −22.66% value delta in redlined areas and widening gap 80% → 121.5% across 76 years; shipped publishable maps & methods.

• LLM/NLP & Retrieval — HF Transformers, PEFT (LoRA/QLoRA), prompt engineering, function calling, embeddings/RAG; built evaluation with accuracy/F1/latency & cost tracking.

• Data Engineering / Analyst — pandas, NumPy, scikit-learn, PyTorch, SQL (MySQL, SQL Server); clean ETL, feature stores, profiling, A/B-like backtests, KPI rollups; GitHub Actions CI; AWS/GCP; GIS (ArcGIS Pro/QGIS).

• Open to: Data Scientist · AI Engineer · Data Analyst · ML Engineer

• Location: Hyattsville, MD · Email: asheermogal786@gmail.com · LinkedIn: https://www.linkedin.com/in/asheermogal/

• Parcel-Equity-Lubbock — Lubbock parcel values + redlining (GIS + notebooks)
• Invoice-Processing-System — OCR → JSON invoice extraction with LLM fine-tuning
• Dr-AI — LLM-powered symptom triage & guidance (Node/Python)
• GPT-Generative-Pre-Trained-Transformer-Project — Intro/visualization of GPT concepts
• ReliefChain — Disaster-relief supply chain optimizer (Python, data pipelines)
• Publications — Papers, writeups, and academic notes