https://dhruvsaikia.com/tags/data/DataHugoBlox Kit (https://hugoblox.com)en-usThu, 15 Jan 2026 00:00:00 +0000https://dhruvsaikia.com/media/icon_hu_f7f9cb5c139bd8fc.pnghttps://dhruvsaikia.com/tags/data/https://dhruvsaikia.com/projects/gdelt-project/Thu, 15 Jan 2026 00:00:00 +0000https://dhruvsaikia.com/projects/gdelt-project/<h2 id="what-i-built">What I built</h2> <ul> <li>Analyzed subsets of a 2PB dataset to identify global risk trends and support stability forecasting.</li> <li>Developed and tuned a Random Forest model and validated it against historical logs.</li> <li>Supported automated anomaly detection and data-driven recommendations.</li> </ul> <h2 id="tools--methods">Tools &amp; methods</h2> <p>Python, SQL, Random Forest, statistical validation, anomaly detection</p> <h2 id="outcome">Outcome</h2> <p>A forecasting + anomaly detection workflow designed to support stakeholder-ready insights from extremely large datasets.</p> <hr>https://dhruvsaikia.com/projects/translink-transit/Mon, 15 Dec 2025 00:00:00 +0000https://dhruvsaikia.com/projects/translink-transit/<h2 id="what-i-built">What I built</h2> <ul> <li>Architected scalable ETL pipelines using PySpark + Spark SQL to ingest 23GB of logs into a Data Lake.</li> <li>Implemented transformations across bronze/silver/gold layers to ensure consistency and data quality.</li> <li>Built interactive Tableau and Power BI dashboards for KPI tracking and performance trends.</li> </ul> <h2 id="tools--methods">Tools &amp; methods</h2> <p>PySpark, Spark SQL, Data Lake, Medallion Architecture, Tableau, Power BI</p> <h2 id="outcome">Outcome</h2> <p>A clean pipeline-to-dashboard workflow that enables business-facing KPI visibility and faster decision-making.</p> <hr>https://dhruvsaikia.com/projects/steam200k-recommender/Sat, 15 Nov 2025 00:00:00 +0000https://dhruvsaikia.com/projects/steam200k-recommender/<h2 id="overview">Overview</h2> <p>This project builds a Top-N game recommendation tool using the Steam-200k dataset. It models user-game interactions and recommends games based on similarity between player profiles.</p> <h2 id="what-i-built">What I built</h2> <ul> <li>Implemented a recommender pipeline using implicit-feedback ALS (Alternating Least Squares).</li> <li>Transformed gameplay behavior into user preference signals to create user profiles.</li> <li>Generated Top-N recommendations by identifying players with similar profiles and surfacing games they play that the user hasn’t seen.</li> </ul> <h2 id="approach">Approach</h2> <ul> <li>Treat play behavior as implicit feedback rather than explicit ratings.</li> <li>Learn latent factors for users and games using ALS.</li> <li>Recommend games with the highest predicted relevance for each user.</li> </ul> <h2 id="tools--tech">Tools &amp; tech</h2> <p>Python, implicit ALS, matrix factorization, data preprocessing, evaluation/validation</p> <h2 id="outcome">Outcome</h2> <p>A working recommendation tool that produces personalized Top-N suggestions from large-scale interaction data.</p> <hr>