Nilesh Sarkar / Academic Contributions

Research & Curriculum Development at Dayananda Sagar University

LLM Engineering Curriculum

During my undergraduate studies at Dayananda Sagar University, I played a key role in designing the curriculum, lab manuals, and coursework for the LLM Engineering syllabus. This initiative involved structuring a comprehensive learning path that bridges theoretical fundamentals with practical implementation, enabling students to gain hands-on experience in building, fine-tuning, and deploying large language models.

Course Objective: This curriculum provides a structured path from basic programming concepts to production-ready AI applications. Students build real-world projects including web scrapers, chatbots, RAG systems, custom models, and cloud deployments, developing both theoretical understanding and practical implementation skills.

Course Structure

The curriculum is organized as a progressive 9-week journey, with each week introducing new concepts while building on previous knowledge. Students complete hands-on projects that demonstrate mastery of both fundamental principles and advanced techniques in language model engineering.

Week 1 Web AI Researcher
Introduction to language models through practical web scraping and analysis. Students learn to interact with Google Gemini AI, implement robust web scraping pipelines, and apply cosine similarity for semantic analysis. The project culminates in building an AI-powered research assistant capable of summarizing web content with comprehensive error handling.
Topics: Gemini API Integration · Web Scraping · Cosine Similarity · Error Handling
Week 2 — Stateful Chat Systems
Deep dive into recurrent neural networks and sequence modeling. Students implement RNN/LSTM architectures from scratch, build interactive interfaces using Gradio, and explore function calling capabilities. The week's project involves creating a context-aware airline assistance chatbot that maintains conversation state across multiple interactions.
Topics: RNN/LSTM Architectures · Gradio UI Development · Tools and Function Calling · Conversation State Management
Week 3 — Meeting Extractor
Understanding attention mechanisms and tokenization strategies. Students explore the transformer attention mechanism, implement Byte-Pair Encoding (BPE) tokenization, and design structured user interfaces for information extraction. The project focuses on building a system that extracts actionable insights from meeting transcripts using BPE-tokenized input streams.
Topics: Attention Mechanisms · BPE Tokenization · Structured Output · Information Extraction
Week 4 — Prompt Engineering
Comprehensive study of prompt engineering techniques and AI safety. Students master system prompting, learn to steer model behavior effectively, and understand critical security concerns including prompt injection attacks. This week emphasizes responsible AI development and robust prompt design for production systems.
Topics: System Prompting · Model Steering · AI Safety · Prompt Injection Prevention
Week 5 — Semantic RAG
Introduction to Retrieval-Augmented Generation systems. Students learn various document chunking strategies, implement vector databases using ChromaDB and FAISS, and deploy persistent searchable knowledge bases. The project demonstrates how to build scalable RAG systems that efficiently retrieve and utilize external knowledge.
Topics: Chunking Strategies · Vector Databases · ChromaDB/FAISS · Persistent Storage · Semantic Search
Week 6 — Local AI Engine
Running large language models locally without internet dependency. Students work with Ollama for local model deployment, understand quantization techniques for model compression, and create custom Modelfiles. This week emphasizes privacy-preserving AI and edge deployment scenarios where models operate entirely offline.
Topics: Ollama · Model Quantization · Modelfiles · Edge Deployment · Privacy-Preserving AI
Week 7 — Agentic Reasoning
Advanced exploration of autonomous AI agents. Students implement self-reflective agents using LangGraph, design agent workflows with decision-making loops, and build systems capable of planning, acting, and learning from feedback. The project showcases agents that can reason about their own outputs and iteratively improve their responses.
Topics: LangGraph · Self-Reflective Agents · Agent Workflows · Reasoning Loops · Autonomous Decision-Making
Week 8 — Model Judge & Eval
Systematic evaluation and monitoring of AI systems. Students implement telemetry for cost and token tracking, learn RAG evaluation using the Triad framework (context relevance, groundedness, answer relevance), and quantify model performance metrics. This week focuses on measuring and optimizing the ROI of AI applications in production environments.
Topics: Telemetry and Monitoring · Cost/Token Tracking · RAG Triad Evaluation · Performance Metrics · ROI Analysis
Week 9 — Production Capstone
Culminating project deploying a production-ready autonomous researcher agent. Students build complete applications using FastAPI, implement Human-in-the-Loop (HITL) workflows for quality assurance, and handle deployment considerations including API design, error handling, and scalability. The capstone integrates all previous concepts into a fully functional, deployable AI system.
Topics: FastAPI · Production Deployment · Human-in-the-Loop · API Design · System Integration

Learning Resources

The complete curriculum includes Jupyter notebooks for each week, comprehensive PDF lab manuals with step-by-step instructions, and a compiled reference document containing all course materials. Students also receive setup scripts, environment configurations, and detailed troubleshooting guides.

Course Credits

Prepared By:
Nilesh Sarkar, 5th Semester Student
Department of Artificial Intelligence and Robotics Engineering
Dayananda Sagar University, Bangalore

Under the Guidance of:
Dr. Pramod Kumar Naik, Chairperson
Department of Artificial Intelligence and Robotics Engineering
Dayananda Sagar University, Bangalore