Generative AI
Internship Report
M. NIHARIKA | 22501A1273
Internship Program Overview
Program Partners Primary Goal Learning Structure
A 10-week virtual internship provided To equip students with essential Comprised of multiple learning
by AICTE, Google Cloud, and knowledge and hands-on experience pathways, hands-on labs, and
EduSkills. in Generative AI and Google Cloud real-world use cases on the Google
tools. Cloud Platform.
Learning Path 1: Introduction to Generative AI
This foundational path covered the core concepts, tools, and ethical
considerations of Generative AI.
✓ What is Generative AI?
✓ Introduction to Large Language Models (LLMs)
✓ Introduction to Responsible AI
✓ Prompt Design in Vertex AI
✓ Applying AI Principles with Google Cloud
Key Concept: Responsible AI
Key Principles Google Cloud Tools
Focused on ensuring AI systems are ethical, fair, and Gained experience with tools to build trustworthy AI. This
beneficial to society. The key principles studied include: included using the What-If Tool to check for model bias and
Fairness & Bias Reduction AutoML features to explain model predictions.
✓
✓ Interpretability & Transparency
✓ Accountability
✓ Privacy & Safety
Learning Path 2: Gemini for Google Cloud
This path explored how Gemini, Google's advanced multimodal model, empowers different roles across the tech
landscape.
For Application Developers: Assisting in all phases of the SDLC, from strategy and brainstorming to
automated code generation.
For Cloud Architects: Automating resource provisioning, optimizing costs, and aiding in strategic capacity
planning.
For Data Scientists: Aiding in data preparation, feature engineering, and accelerating model development.
For Security Engineers: Enhancing threat intelligence, automating vulnerability assessments, and
streamlining incident response.
Learning Path 3: Advanced
GenAI
This path explored the core architecture and practical application of
the models that power modern Generative AI.
→ Encoder-Decoder Architecture
→ Attention Mechanism
→ Transformers and BERT Models
→ Creating Image Captioning Models
→ Vector Search and Embeddings
→ Machine Learning Operations (MLOps)
Core Platform: Vertex AI
Unified ML Platform Key Features Used
Vertex AI is Google Cloud's managed platform to train, Gained hands-on experience with several key Vertex AI
deploy, and scale ML models. We used Generative AI Studio services, including PaLM APIs for text generation, Vector
within Vertex AI to design, test, and refine prompts for large Search for efficient similarity matching, and MLOps
language models. practices for pipeline deployment.
Advanced Topic: Image Generation
Explored models designed to create new, original visual content from
text prompts and other inputs.
✓ Studied core models like GANs (Generative Adversarial
Networks) and VAEs.
✓ Learned the Encoder-Decoder architecture used to bridge
computer vision and language.
✓ Practiced building image captioning models to automatically
describe visual content.
Advanced Topic: Gemini & Multimodality
Gemini's Multimodal Power
Studied how Gemini processes and understands complex documents
containing more than just text. This includes interpreting images,
charts, and tables within a single prompt.
Explored Multimodal RAG (Retrieval-Augmented Generation) to pull
external knowledge to provide a more complete, context-aware
response.
Summary of Skills Acquired
Core GenAI Concepts Google Cloud Platform Responsible AI
LLMs, Prompt Engineering, Vertex AI Studio, Gemini API, MLOps, Hands-on application of Fairness &
Transformers, Attention Mechanism, Vector Search, BigQuery, and Cloud Bias detection, Interpretability &
and Encoder-Decoder Architecture. Storage. Transparency, and Privacy & Safety.
Conclusion
Google Cloud Generative AI is transforming technology by enabling
the creation of new content at scale, moving AI from a specialized tool
to an accessible platform for innovation.
The platform's strong focus on responsibility, ethics, and transparency
is not just a feature, but a core component. This sets a strong
foundation for building ethical and inclusive AI solutions for the future.
Thank You
Questions?
M. Niharika | 22501A1273