Featured Work
Data Systems Built for Production
From automated pipelines and real-time processing to deployed platforms — each system is engineered end-to-end and runs in production.
Personal Project • Full-Stack
EchoMap — Mapping the News
Fully automated data pipeline that collects, classifies, and maps news events in real time.
Uses Telegram API for data ingestion, Gemini AI for entity extraction and geolocation,
and PostGIS for spatial queries. Features multilingual support and event categorization
(protests, disasters, conflicts).
Python
Telegram API
Gemini AI
PostGIS
React
Leaflet
Docker
CapRover
Data Engineering • Analytics
E-Commerce Sales Intelligence
End-to-end e-commerce analytics platform: automated Selenium scraper collects product data on a 6-hour schedule,
stores it in PostgreSQL with historical tracking. Interactive Dash dashboard shows KPIs, stock-out rates,
high-demand products, price distributions, and trends — all auto-refreshing.
Python
Selenium
Dash / Plotly
PostgreSQL
Docker
CapRover
Machine Learning • NLP
Job Application Analyzer
NLP-powered tool that scrapes LinkedIn jobs via Playwright, computes resume-to-job semantic similarity
using SentenceTransformer embeddings, and runs statistical analysis (t-test, Mann-Whitney, Cohen's d,
chi-square) to validate whether similarity predicts interview success.
Python
SentenceTransformers
Playwright
scikit-learn
PostgreSQL
SciPy
More Work
Analytical Case Studies
Engineering-focused case studies demonstrating pipeline design, statistical analysis, ML pipelines, and SQL mastery.
Onboarding Funnel & A/B Testing
Mobile App Analytics
Analyzed mobile app onboarding funnel (5 stages) and evaluated A/B experiments with 95% confidence intervals.
Discovered ~49.8% lift in one experiment variant. Segmented by funnel types and gender.
Python
pandas
SciPy
A/B Testing
IMU Sensor Analysis
Automotive / Dashcam
Built ML pipeline for dashcam IMU data: feature extraction, RandomForest inference, and event detection.
Designed stop-sign compliance annotation protocol with structured classification categories.
Python
scikit-learn
RandomForest
Signal Processing
Campaign & Partner Analytics
Ad Tech / Marketing
Partner attribution analysis, conversion funnels, and profitability modeling for mobile ad campaigns.
Identified underperforming partners and calculated loss-per-install metrics across countries.
Python
pandas
PostgreSQL
SQLAlchemy
Inventory SQL & Star Schema
E-Commerce / Streaming
Advanced SQL analysis: max sales between restocks using window functions, stock level calculations.
Designed a star schema for audio streaming events with fact and dimension tables.
SQL
Snowflake
Window Functions
Data Modeling
Multi-Device & VPN Detection
Cybersecurity / Identity
Classified users as single-device vs multi-device using device IDs, OS fingerprints, and IP overlap.
Built VPN detection using Haversine distance and impossible-speed analysis (>1200 km/h).
Python
pandas
Geospatial
Haversine