Harsha Sathish

AI Intern @ BMW Group | M.Sc. AI @ THWS, Germany | Ex GE Vernova · Infosys

Deep Learning · NLP · Computer Vision · RAG Systems · LLMs · Active Learning · Medical Imaging · Reinforcement Learning · Python · Java · Salesforce · Agentic AI

Get In Touch View Projects

Certifications

IBM RAG and Agentic AI: Build Next-Gen AI Systems

IBM & Coursera

November 2025

Verify Certificate

Salesforce Certified AI Associate

Salesforce

2024

Technical Skills

Python

Advanced

Machine Learning

Advanced

RAG & LLMs

Advanced

Computer Vision

Intermediate

NLP

Intermediate

Deep Learning

Intermediate

Salesforce

Advanced

Java / Apex

Advanced

Professional Experience

AI Intern

BMW Group, Munich, Germany

2025 – Present

  • Working in the QMT Front Desk Lab on AI-driven quality management solutions
  • Attended BMW i3 (Neue Klasse) world premiere as part of the AI team

HiWi / Research Assistant

Technische Hochschule Würzburg-Schweinfurt, Germany

November 2024 – March 2025

  • RAG Developer for the Bearings / Mechatronics Lab
  • Built document processing pipeline with local LLM deployment
  • Developed semantic retrieval system for engineering knowledge Q&A

Senior Associate Consultant

Infosys, Trivandrum, India

June 2024 – February 2025

  • Developed critical Salesforce requirements for GE Vernova Business Units
  • Worked with Apex, Aura Components, LWC, Flows, SOQL/SOSL
  • Obtained Salesforce Certified AI Associate Certification
  • Led team optimization of Salesforce Architecture

Software Engineering Specialist & DT Intern

GE Vernova, Bangalore, India

January 2023 – May 2024

  • Led critical data migration project with $250,000 annual cost savings
  • Built and customized Salesforce solutions using Apex, LWC, Flows
  • Customized Flosum CI/CD tool for streamlined deployment

Research Intern

Litmus7, Ernakulam (Remote), India

July 2022 – November 2022

  • Developed Streamlit application for data preprocessing
  • Built classification models using Python, TensorFlow, and ML libraries

Featured Projects

H. Pylori Bacteria Detection on WSI Images Ongoing

Active learning pipeline (YOLOv8 + Faster R-CNN) for automated detection of H. pylori in veterinary gastric biopsy whole slide images, in collaboration with LABOKLIN GmbH. Dataset grew from 1,153 to 2,412 verified positive annotations across 3 active learning cycles, improving precision from ~28% to 80%.

PythonYOLOv8Faster R-CNNActive LearningMedical Imaging

CAIRA — MAI Student Assistant Ongoing

Conversational AI assistant built for THWS Master of Artificial Intelligence students. Covers admissions, housing, courses, visa procedures, and campus life — helping international students navigate their academic journey in Germany.

PythonRAGLLMNLPChatbot

Bearings.THWS — RAG Document Pipeline

Built a document processing RAG system for the THWS Bearings / Mechatronics Lab. Designed multi-source ingestion, semantic retrieval, and local LLM deployment for engineering knowledge Q&A.

PythonRAGDocument ProcessingLocal LLMNLP

Sign Language Recognition System

Deep learning model using ResNet18 achieving 87.49% accuracy on ASL alphabet recognition. Implemented CLAHE preprocessing, morphological operations, and test-time augmentation for 22-class classification.

PythonResNet18PyTorchOpenCVComputer Vision

Multi-Object Trajectory Tracking

Real-time tracking system using particle filters and the Hungarian algorithm for data association. Handles 2000+ particles with DBSCAN clustering and systematic resampling, robust to sensor noise and dropouts.

PythonParticle FilterDBSCANHungarian Algorithm

BlackJack Strategy Optimization

Q-Learning and Double Q-Learning agents trained over 500,000 episodes achieving a 43.5% win rate vs. a 27% random baseline, using epsilon-greedy exploration and systematic state-space coverage.

PythonQ-LearningReinforcement LearningGame Theory

Publications

A Comparative Study on the Working of GNN and CNN on Panoramic X-Rays in Prediction of Dental Diseases

Analyzed deep learning architectures for dental image analysis, comparing Graph Neural Networks and Convolutional Neural Networks for medical diagnosis applications. IEEE Conference (Scopus Indexed).

A Strategy for Detecting Malicious Spam Emails using various Classifiers

Developed a comprehensive strategy for spam email detection using multiple classification algorithms, implementing comparative analysis of various machine learning approaches. IEEE Conference (Scopus Indexed).

Aiding the Visually Impaired using Artificial Intelligence and Speech Recognition Technology

Research on accessibility technologies leveraging computer vision and speech technology to assist visually impaired individuals in daily activities. IEEE Conference (Scopus Indexed).

Honors & Awards

Technology Excellence CIO Awards

Twice nominated for outstanding technical contributions at GE Vernova

Deliver with Focus Award

Recognized for high impact performance and project delivery excellence

Applause Award

Multiple recognitions for exceptional work quality and team collaboration

Restaurant Co-founder

Co-founded a restaurant and managed two cafeterias, implementing POS and online ordering systems

Let's Connect

Currently AI Intern at BMW Group while pursuing M.Sc. AI at THWS Germany.
Open to conversations about RAG, LLMs, AI research, and hackathon collaborations.

harshasathish2001@gmail.com
+49 152 10 692 384
Munich, Germany

Languages: English (C1) • German (A2, working toward B1) • Tamil (Native) • Malayalam (Native) • Hindi (Intermediate)