Applied AI • NLP • Model Evaluation

Data Scientist building and evaluating practical AI systems

Building and evaluating NLP classifiers and RAG retrieval systems with Python, scikit-learn, PyTorch, Transformers, and rigorous evaluation.

LinkedIn GitHub Resume

Email me

Featured Applied AI & Data Science

Financial Complaint Auto-Routing NLP workflow showing leakage-safe preparation, TF-IDF and DistilBERT evaluation, selective routing, and retrospective validation.

NLP CLASSIFICATION / MODEL EVALUATION

COMPLETED V1–V2 STUDY

Financial Complaint Auto-Routing with NLP

Leakage-safe eight-class CFPB text-classification study comparing a locked TF-IDF + Linear SVM benchmark with a frozen DistilBERT challenger and human-in-the-loop routing.

Removed duplicate-text leakage and used group-aware 2024 development/test splits with zero normalized-text overlap.
On the shared 2024 benchmark, the frozen DistilBERT challenger increased Macro F1 from 0.7671 to 0.7949; both frozen models were later compared on a 30,156-row retrospective 2025 cohort.

Pythonscikit-learnTF-IDFLinear SVMPyTorchDistilBERTModel Evaluation

GitHub Repo Case Study Architecture

CivicLens hybrid RAG workflow showing NYC 311 document ingestion, pgvector retrieval, cited answers, Next.js UI, analytics routing, and local evaluation.

APPLIED AI / RAG EVALUATION

HOSTED PORTFOLIO DEMO

CivicLens RAG — NYC 311 Operations Copilot

Hosted hybrid RAG application for grounded NYC 311 documentation Q&A with semantic + full-text retrieval, validated citations, safe abstention, and bounded analytics.

Combined semantic search and PostgreSQL full-text retrieval with deterministic Reciprocal Rank Fusion (RRF) over curated NYC 311 knowledge.
In the approved real-local evaluation, hybrid retrieval reached 83.9% Recall@5 and 92.9% expected-source retrieval across 14 retrieval-eligible questions.

PythonHybrid RetrievalRRFPostgreSQLpgvectorFastAPIRAG Evaluation

Live Demo Case Study GitHub Repo Architecture

Data & Cloud Foundations

AZURE DATA ENGINEERING

NYC 311 Service Requests Lakehouse

Azure-first medallion lakehouse for NYC 311 operational analytics, transforming raw API data into analytics-ready bronze, silver, and gold datasets.

Azure Data FactoryADLS Gen2DatabricksPySparkDelta LakePythonSQLPower BIGitHub Actions

Case Study GitHub Repo Architecture Proof

REAL AWS CLOUD PROOF

Cloud Flight Fare Pipeline

Real AWS cloud proof project showing EventBridge Scheduler -> ECS/Fargate -> Flight API -> S3 Bronze -> Redshift Serverless -> dbt staging/marts/tests -> CloudWatch Logs, with proof screenshots, runbooks, and cost/secret safety notes.

AWSECS/FargateEventBridgeS3RedshiftdbtDockerCloudWatch

Case Study GitHub Repo Architecture Proof

PYTHON DATA INGESTION

Travelpayouts Flight Collector

Python API ingestion project that collects live Travelpayouts flight fare data and publishes dated CSV snapshots for analytics.

PythonAPI IngestionCSVSchedulingpytestGitHub Actions

GitHub Repo Data

Supporting Analytics & Applications

DATA SCIENCE

Flight Price Analytics

Time-based Buy/Wait classification and fare forecasting using route-level price snapshots, feature engineering, and backtesting.

Pythonscikit-learnForecastingTime-Based SplitPower BI

GitHub Repo ML Pipeline

DATA ANALYTICS

Floral Daily SKU Analysis

Daily SKU sales, margin, promotion, holiday, and waste analysis using Python, SQL, and Power BI to support inventory and business decisions.

PythonSQLPower BIInventory AnalyticsTrend Analysis

GitHub Repo SQL Analysis

AI REPORTING SAAS

Sumryze - AI-Powered SEO Reporting Dashboard

SaaS-style dashboard for automated SEO reporting, AI-generated summaries, analytics visualizations, and client-ready insights.

Next.jsTypeScriptTailwindOpenAIREST APIsVercel

GitHub Repo Live Site

Skills & Tools

Programming & Analysis

PythonSQLpandasNumPyStatistical AnalysisExploratory Data Analysis

Machine Learning & NLP

scikit-learnTF-IDFLinear SVMPyTorchTransformersDistilBERTText Classification

Model Evaluation & Experimentation

Macro F1CoverageModel ComparisonError AnalysisLeakage PreventionGroup-Aware ValidationTime-Based ValidationBacktesting

AI / RAG

EmbeddingsVector SearchHybrid RetrievalRRFPostgreSQL/pgvectorRetrieval EvaluationGrounded GenerationSafe Abstention

Data & Analytics

PostgreSQLPySparkDelta LakedbtDatabricksPower BIDimensional Modeling

Engineering & Cloud

FastAPIDockerGitHub ActionsAzureAWSApache AirflowCloud Data Pipelines

About

I’m a Data Scientist focused on applied AI and practical AI systems. I build and evaluate NLP classifiers, retrieval systems, and RAG applications, with particular attention to model quality, failure modes, grounded outputs, and reliable evaluation.

I’m currently completing an M.S. in Information Technology with a Data Analytics concentration, with coursework focused heavily on data science, machine learning, predictive analytics, and statistical modeling.

My background in web development and cloud data engineering across AWS and Azure helps me connect the full workflow—from data ingestion and transformation to models, APIs, and user-facing applications.

Contact

Interested in collaborating on data science, applied AI, or portfolio projects? Reach out and I will follow up.

Email me LinkedIn GitHub