| 🤖 Data Scientist | 🛠️ MLOps Engineer | 📍 Geospatial Analyst |
|---|---|---|
| Building production ML that classifies, detects and forecasts at scale | Automating CI/CD, secret management and scheduled jobs on GCP | Turning raw vehicle telematics into location intelligence |
| 🎓 Amrita Vishwa Vidyapeetham | 🚛 VE Commercial Vehicles Ltd. | 📍 Coimbatore, India |
|---|---|---|
| M.Sc. Data Science | Data Scientist – I | Open to Data Scientist & Data Analyst roles |
I own end-to-end ML systems at VECV, from raw telematics data to monitored production models. I start with the business decision, pick the simplest model that holds up in the field, and build monitoring so it stays reliable after launch. Production deployments, not just notebooks.
| 100K+ | 98–99% | 500 m | 99.9% |
|---|---|---|---|
| Fuel sensors classified monthly | Trip detection accuracy in customer-validated pilots | Hotspot precision for vehicle stoppage locations | Uptime on automated 15-minute data jobs |
- Fuel Sensor Classification (GCP): End-to-end pipeline on BigQuery telematics data using breakpoint detection and rule-based confidence gating. It replaced 100% manual inspection and runs autonomously every month.
- Trip Detection for Mining & Construction: Core algorithm behind trip counts on the Control Tower platform, with 98–99% accuracy despite GPS noise and rugged terrain. Includes an efficacy-monitoring framework and earned an Appreciation Letter from the Executive Vice President.
- Geospatial Hotspot Clustering: BIRCH + Haversine hierarchical clustering with boundary-aware refinement. It scales across India without proportional infrastructure growth.
- MLOps & CI/CD: GitHub Actions + GCP pipeline with automated quality checks, unit tests, Gitleaks secret scanning, Secret Manager and IAP-based VM deployment. Zero committed credentials.
- Earlier work: E-commerce sell-out forecasting models and Power BI automation that cut reporting time by 30%.
| Project | Description | Stack |
|---|---|---|
| Geospatial-Hotspot-Clustering | Hybrid spatial clustering pipeline that finds repeated geographic hotspots in large-scale vehicle event data, built for datasets that exceed the scalability limits of standard clustering | Python, BIRCH, Haversine, Hierarchical Clustering |
| gcloud_CI-CD | Automated CI/CD pipeline on GCP with secret scanning, secure configuration delivery and scheduled data-processing jobs | Python, GitHub Actions, GCP, Gitleaks |
Languages & Data
Machine Learning
Cloud & MLOps
Analytics & Visualization
I'm open to Data Scientist and Data Analyst roles. If you're building ML systems that need to work in the real world, I'd love to talk.
