Puru Singh

Bangalore, Karnataka, India

Phone: +91 7838014645

Email: puru.singh@outlook.in

GitHub Icon LinkedIn Icon Download Resume

Skills & Certifications

Skills:

  • Python
  • PySpark
  • Apache Spark
  • SQL
  • Databricks
  • Apache Airflow
  • Machine Learning
  • Generative AI - LangChain, RAGs, Ollama
  • C++
  • Power BI
  • Kotlin

Certifications:

Languages: English, Hindi

Typing Speed: 172 WPM

Education

Graphic Era Deemed to be University, Dehradun, UK

B.Tech in Computer Science (2020 - 2024), GPA: 8.73

Experience

Genpact, Data Engineer - Business Analyst — Bangalore, Karnataka — Oct 2024 – Present

  • Developed, Migrated and optimized complex ETL jobs and pipelines from PostgreSQL based Greenplum to SparkSQL on Databricks.
  • Improved performance by up to 80% using through logical flow enhancement and Spark-specific optimizations.
  • Designed and orchestrated scalable workflows and connection modules using Python for Databricks and Apache Airflow.

Genpact, Intern - Data Analyst - Gen AI — Hyderabad, Telangana — Feb 2024 – Jul 2024

  • Built a Gen AI enabled financial summary generator for 10-K reports using Python - LangChain - Ollama.
  • Created a Power BI dashboard for consumer banking insights.

DRDO, Intern - Android Development and Security — New Delhi, Delhi — Jul 2023 – Aug 2023

  • Built a Kotlin app for SMS-based notification forwarding.
  • Performed mobile security audits using MobSF.

Projects

Loocle – A Self-Hosted Photo Backup and Web App

  • Self-hosted photo management system replicating Google Photos’ backup and viewing experience, running entirely on local devices without any cloud dependencies.
  • Automated Backups from Androids using Wireless ADB
  • Auto Generated Tags using locally running LLMs - Ollama - Gemma3
  • Face Recognition using Multi-Shot classification, allowing searching through faces.
  • Front-End using Flutter for cross platform support.
  • Zero reliance on third-party cloud services.

Face Recognition using One-Shot Classification

  • Learn and Recognise faces using a single source image and embedding difference.
  • Over 90% speedup over traditional face recognition in both Learning and Recognition
  • Used FaceNet + Haar-Cascade, and one-shot classification.

Achievements

  • Top 50 – Coursera Employment Hackathon (2023)
  • 3rd Place – Intel OneAPI Hackathon, IIT Roorkee (2023)