Systems & Quantitative Analytics Portfolio

A collection of data science projects focused on complex systems, spatial-temporal modeling, quantitative finance, and
computational linguistics.

Rather than standard machine learning classification tasks, these projects emphasize backend database architecture,
formal statistical modeling, dynamic systems, and interactive analytical dashboards.

---

Portfolio Architecture & Tech Stack

All four projects are built on a unified core pipeline, using domain-specific libraries to address distinct analytical
problems:

$$\text{PostgreSQL} \longrightarrow \text{Python (Pandas / SciPy)} \longrightarrow \text{Statistical & Algorithmic Modeling} \longrightarrow \text{Interactive Dashboards}$$

Project	Primary Domain	Core Algorithms & Methods	Key Libraries
1. Urban Mobility	Spatial Data Engineering	Moran's I, DBSCAN, Time-Series Forecasting	GeoPandas, PySAL, Prophet
2. Financial Risk	Quantitative Finance	GARCH(1,1), Monte Carlo, Value at Risk (VaR), PCA	SciPy, statsmodels, arch
3. Linguistics	Computational NLP	Word2Vec, Cosine Drift, KL Divergence, Zipf's Law	spaCy, Gensim, NLTK
4. Agent Simulation	Complex Systems	Mesa Agent-Based Framework, Sensitivity Analysis	Mesa, NetworkX, NumPy

---

Projects Overview

1. Urban Mobility & Transit Reliability

• Objective: Analyze public transit data to understand spatio-temporal bottlenecks and schedule reliability.
• Key Findings: Identified spatial autocorrelation in delay clusters using Moran's I and isolated key route
  bottlenecks during peak congestion hours.
• Pipeline Highlights:
  • PostGIS spatial indexing (GIST) for rapid geographic filtering.
  • Time-series delay forecasting using Prophet conditioned on weather and temporal features.
  • Interactive route choropleth maps built with Folium/Plotly.

2. Financial Markets & Risk Analytics

• Objective: Model asset volatility, evaluate portfolio risk metrics, and project stress-test scenarios.
• Key Findings: Constructed a dynamic risk model showing time-varying volatility dynamics and portfolio drawdowns
  under simulated tail-risk conditions.
• Pipeline Highlights:
  • Time-series PostgreSQL schemas leveraging SQL window functions (LAG, LEAD).
  • GARCH(1,1) volatility clustering and 10,000-path Monte Carlo simulations.
  • Interactive portfolio risk dashboards tracking parametric vs. historical Value at Risk (VaR).

3. Computational Linguistics (Historical Corpus Analysis)

• Objective: Measure semantic drift, vocabulary shifts, and syntactic changes across historical text corpora.
• Key Findings: Quantified directional shift vectors for target terminology across a century of text data using
  distributional semantics.
• Pipeline Highlights:
  • Full-text search GIN indexes in PostgreSQL for corpus storage.
  • Word2Vec model alignment to calculate cosine similarity shifts over time.
  • Dimensionality reduction using UMAP/t-SNE to map semantic trajectories.

4. Agent-Based Simulation (Dynamic Systems)

• Objective: Simulate emergent macro-level behaviors from micro-level agent interactions.
• Key Findings: Mapped phase transitions and tipping points where small shifts in individual behavior rules yield
  system-wide equilibrium shifts.
• Pipeline Highlights:
  • Object-oriented simulation models constructed with Mesa.
  • Batch logging of tick-by-tick simulation states directly into PostgreSQL.
  • Network topology analysis using NetworkX alongside animated spatial visualizations.

---

Repository Structure

├── 01_urban_mobility/
│   ├── 01_data_ingestion.ipynb
│   ├── 02_spatial_stats.ipynb
│   ├── 03_predictive_modeling.ipynb
│   └── 04_visualization_dashboard.ipynb
├── 02_financial_risk/
├── 03_computational_linguistics/
├── 04_agent_simulation/
├── src/                        # Modular helper functions across projects
│   ├── db_utils.py             # PostgreSQL connection and query wrappers
│   ├── stats_utils.py          # Custom statistical functions
│   └── visualization.py        # Shared Plotly/Matplotlib formatting
├── environment.yml             # Conda environment specifications
└── README.md