MSc Data Analytics @ BSBI Berlin | Data Scientist | ML Engineer
Architecting Scalable Forecasting Engine & Industrial Computer Vision
🎯 Actively Interviewing: Seeking Full-Time Data Science / ML Engineer roles.
📍 Location: Berlin, Germany (Hybrid/Remote) & Open to GCC Roles in Hyderabad, India.
📅 Availability: Immediate Start (April 2026).
🇩🇪 Language: German (A2 Elementary – Advancing to B1).
- 🏗️ Big Data Architect: Engineered a 15.2M record pipeline using PySpark and GCP Dataproc.
- 📉 Memory Optimization: Achieved 70% RAM reduction via advanced data downcasting and feature engineering.
- 👁️ Computer Vision: Developed an industrial-grade ReUNet model achieving 91.7% Accuracy.
- 👥 Technical Leadership: Former Founder & Lead Developer at CodeMacrocosm; scaled an open-source community to 2,100+ members and 1.2k+ Stars.
| Category | Tools & Technologies |
|---|---|
| Machine Learning | Python, Scikit-Learn, LightGBM, XGBoost, TensorFlow, OpenCV |
| Data Engineering | PySpark, GCP (Dataproc/BigQuery), SQL (PostgreSQL), ETL Pipelines |
| Statistical Research | Predictive Modeling, Time-Series Forecasting, Sales Analytics |
| Software Ops | Git/GitHub, DSA (O(n) Optimization), Docker, Flask API |
Scale: 15.2 Million Transactions | Tech: PySpark, LightGBM, GCP
- The Problem: High-latency and memory crashes during large-scale retail demand forecasting.
- The Solution: Implemented a Tweedie-loss LightGBM model with a memory-optimized data loader.
- The Result: 70% less memory usage and 15% higher accuracy than baseline models. View Project →
I write Production-Grade Python. I focus on
+@ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @+
@@ o o @@
@@ | | @@
@@ _L_L_ @@
@@ ❮\/__-__\/❯ Programming isn't about what you know @@
@@ ❮(|~o.o~|)❯ It's about what you can figure out @@
@@ ❮/ \`-'/ \❯ @@
@@ _/`U'\_ @@
@@ ( . . ) .----------------------------. @@
@@ / / \ \ | while( ! (succed=try() ) ) | @@
@@ \ | , | / '----------------------------' @@
@@ \|=====|/ @@
@@ |_.^._| @@
@@ | |"| | @@
@@ ( ) ( ) Testing leads to failure @@
@@ |_| |_| and failure leads to understanding @@
@@ _.-' _j L_ '-._ @@
@@(___.' '.___) @@
+@ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @ @+
