Traffic Signal Optimization — RL vs MILP
Deep Q-Learning vs Webster's-formula MILP for a 4-way intersection under Indian mixed-traffic — custom simulation with cars, bikes, buses & autos.
I turn complex operational problems into optimized, data-driven decisions — from MILP models scheduling 40+ trains a day to ML pipelines estimating EV battery health.
My work spans forecasting, mathematical optimization, simulation, machine learning and NLP — applied to transportation, supply chains, inventory planning and intelligent systems.
MILP, train scheduling, resource allocation — PuLP, CPLEX & Gurobi at industrial scale.
Dashboards, KPI tracking and business intelligence with Power BI, Tableau & SQL.
OTIF analysis, demand forecasting and inventory planning across FMCG & retail data.
Churn prediction, classification and large-scale text preprocessing pipelines.
SOH estimation, drive-cycle analysis and thermal data modeling for electric vehicles.
Land-use mapping and regional planning with QGIS & OpenStreetMap.
Research and industry roles across optimization, analytics, EV systems and urban planning.
Hyundai-funded EV project: battery State-of-Health estimation using ML models and temperature data; drive-cycle data analysis & statistical framework development.
Sivagangai Regional Plan 2047 (DTCP Tamil Nadu-funded consultancy): land-use mapping and regional development analysis using QGIS and OpenStreetMap.
Built a MILP model for freight-prioritized train scheduling across Chakradharpur Division (SER) — capacity and headway constraints for 40+ daily freight & passenger services.
Analyzed Li-ion battery pack assembly for Bajaj & TVS EV platforms; flagged 3 failure modes in cell-module integration during DFM review.
Built a customer churn prediction model on 50K+ records — 83% accuracy — and visualized key insights via Power BI dashboards.
Every project below shipped with a measurable result — not just a notebook.
Deep Q-Learning vs Webster's-formula MILP for a 4-way intersection under Indian mixed-traffic — custom simulation with cars, bikes, buses & autos.
Freight-prioritized MILP scheduling model for Chakradharpur Division using real Working Time Table data — single-line & junction constraints.
Binary sentiment classifier trained on 1.6M tweets — full pipeline from raw text to TF-IDF to Logistic Regression, deployed as a live Streamlit app.
ARIMA/Prophet time-series forecasting with reorder alerts, backed by Power BI dashboards for inventory & product-level planning.
Discrete-event simulation with routing optimization — worker allocation and layout redesign for measurable throughput gains.
OT%, IF% & OTIF% service-level KPI analysis for AtliQ Mart across 6 product categories — logistics bottlenecks affecting contract renewals.
ILP framework scheduling 50+ tasks across 10 resources with precedence constraints, capacity limits and Gantt visualization.
Scraped & structured the largest US companies by revenue — Requests + BeautifulSoup extraction, Pandas cleaning, production-ready CSV output.
3NF normalized relational schema for 8 entities, with 15+ SQL queries for revenue trends and customer segmentation.
Open to roles in Business Analytics, Operations Research, Supply Chain Analytics & Data Analytics — anywhere the work involves real problem-solving, not just reporting.