AI-GR501 — Machine Learning & Intelligent Systems
Institute: Institute for AI Governance, Cybersecurity & Risk Management (IAIGCRM)
Status: Academic model syllabus — v0.1 editorial draft; not accredited or approved by a degree-granting institution
Format: 15 weeks; proposed 3 semester credits; 3 contact hours/week plus independent work
Academic calendar: Instructor to supply dates, location, office hours, accessibility office, institutional policies and official credit-hour equivalence.
Academic standards and responsible AI policy
Students may use AI tools for formative explanation, brainstorming, debugging, or critique where authorized, but must disclose meaningful assistance, verify claims and citations, protect sensitive data, and independently explain submitted work. Fabricated references, undisclosed generated submissions, and submission of code or reasoning students cannot defend violate course expectations. Instructors must offer an equivalent non-AI route when necessary for accessibility, privacy, or equitable access. Research involving human participants, sensitive data, or live systems requires appropriate approvals. All security exercises use owned or expressly authorized environments.
Inclusive teaching and assessment
Use accessible readings, captioned media, reproducible notebooks or non-programming alternatives as appropriate, formative feedback, transparent rubrics, and documented accommodations. Assess conceptual understanding separately from AI-assisted output quality. A final oral defense or live demonstration tests authorship and comprehension.
General grading principles
Each assessment receives a rubric for technical accuracy, evidence, reasoning, reproducibility, ethical awareness, and communication. Late work, attendance, reassessment, academic integrity, and accessibility procedures must be finalized by the adopting institution.
Editorial notes
All readings are indicative; the instructor must verify editions, licensing, links, publication status, and accessibility before adoption. Course credits and prerequisites are proposals rather than institutional approvals.
Course profile
Level: Graduate technical
Prerequisites: Proficiency in Python, linear algebra, probability, statistics and basic algorithms; bridging modules permitted.
Catalog description: A mathematically and experimentally rigorous study of learning algorithms and intelligent-system engineering, including representation learning, transformer architectures, RAG, fine-tuning, explainability, distribution shift, operational security and post-deployment monitoring.
Measurable learning outcomes
- Derive and compare major learning objectives, regularization approaches and optimization methods.
- Design leakage-resistant experiments and quantify uncertainty, robustness and generalization.
- Implement and evaluate neural architectures and transformer-based components.
- Compare prompting, retrieval, parameter-efficient fine-tuning and full fine-tuning under controlled conditions.
- Engineer and test a provenance-aware RAG or comparable intelligent-system pipeline.
- Document model security, operational monitoring, drift response and reproducibility with defensible release gates.
Assessment and weighting
| Component | Weight |
| Mathematical derivations and problem sets | 20% |
| Five reproducible technical laboratories | 30% |
| Research paper replication and critique | 15% |
| Midterm technical examination | 10% |
| Final system, ablation study, report and defense | 25% |
Total: 100%. The capstone must include an individually attributable contribution, documentation and a defense.
Detailed 15-week instructional plan
Each week specifies measurable objectives, instructional activities, laboratory or seminar requirements, security integration and assessment evidence. Faculty must assign exact chapters and articles before delivery.
Week 01 — Statistical learning theory and empirical risk
Weekly objectives: (1) Explain the principal concepts, assumptions and limitations of statistical learning theory and empirical risk. (2) Apply the week's methods in loss functions, bias-variance, generalization. (3) Engineer reproducible experiments, evaluate adversarial risks and define operational evidence. (4) Interpret evidence and communicate at least one limitation or trade-off.
Instructional plan: Instructor-led conceptual seminar; guided worked example; structured discussion or peer review; independent reading and preparation.
Laboratory / seminar requirements: Loss functions, bias-variance, generalization. A tested implementation, benchmark, experiment report or technical design. Use only synthetic, public, or explicitly approved data; record procedures and limitations.
Security & governance checkpoint: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08; identify at least one relevant control, potential misuse or failure mode, and proportionate safeguard.
Assessment evidence: Submit a dated lab log or seminar brief with objective, methods, results, references, limitations and reflection.
Reading assignment: Relevant selections from the course reading list, assigned and verified by the instructor before the term.
Week 02 — Optimization and regularization
Weekly objectives: (1) Explain the principal concepts, assumptions and limitations of optimization and regularization. (2) Apply the week's methods in gradient methods and controlled comparison. (3) Engineer reproducible experiments, evaluate adversarial risks and define operational evidence. (4) Interpret evidence and communicate at least one limitation or trade-off.
Instructional plan: Instructor-led conceptual seminar; guided worked example; structured discussion or peer review; independent reading and preparation.
Laboratory / seminar requirements: Gradient methods and controlled comparison. A tested implementation, benchmark, experiment report or technical design. Use only synthetic, public, or explicitly approved data; record procedures and limitations.
Security & governance checkpoint: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08; identify at least one relevant control, potential misuse or failure mode, and proportionate safeguard.
Assessment evidence: Submit a dated lab log or seminar brief with objective, methods, results, references, limitations and reflection.
Reading assignment: Relevant selections from the course reading list, assigned and verified by the instructor before the term.
Week 03 — Linear models, trees and ensembles
Weekly objectives: (1) Explain the principal concepts, assumptions and limitations of linear models, trees and ensembles. (2) Apply the week's methods in benchmark and ablation. (3) Engineer reproducible experiments, evaluate adversarial risks and define operational evidence. (4) Interpret evidence and communicate at least one limitation or trade-off.
Instructional plan: Instructor-led conceptual seminar; guided worked example; structured discussion or peer review; independent reading and preparation.
Laboratory / seminar requirements: Benchmark and ablation. A tested implementation, benchmark, experiment report or technical design. Use only synthetic, public, or explicitly approved data; record procedures and limitations.
Security & governance checkpoint: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08; identify at least one relevant control, potential misuse or failure mode, and proportionate safeguard.
Assessment evidence: Submit a dated lab log or seminar brief with objective, methods, results, references, limitations and reflection.
Reading assignment: Relevant selections from the course reading list, assigned and verified by the instructor before the term.
Week 04 — Unsupervised and representation learning
Weekly objectives: (1) Explain the principal concepts, assumptions and limitations of unsupervised and representation learning. (2) Apply the week's methods in clustering and embedding evaluation. (3) Engineer reproducible experiments, evaluate adversarial risks and define operational evidence. (4) Interpret evidence and communicate at least one limitation or trade-off.
Instructional plan: Instructor-led conceptual seminar; guided worked example; structured discussion or peer review; independent reading and preparation.
Laboratory / seminar requirements: Clustering and embedding evaluation. A tested implementation, benchmark, experiment report or technical design. Use only synthetic, public, or explicitly approved data; record procedures and limitations.
Security & governance checkpoint: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08; identify at least one relevant control, potential misuse or failure mode, and proportionate safeguard.
Assessment evidence: Submit a dated lab log or seminar brief with objective, methods, results, references, limitations and reflection.
Reading assignment: Relevant selections from the course reading list, assigned and verified by the instructor before the term.
Week 05 — Neural networks and automatic differentiation
Weekly objectives: (1) Explain the principal concepts, assumptions and limitations of neural networks and automatic differentiation. (2) Apply the week's methods in backpropagation derivation. (3) Engineer reproducible experiments, evaluate adversarial risks and define operational evidence. (4) Interpret evidence and communicate at least one limitation or trade-off.
Instructional plan: Instructor-led conceptual seminar; guided worked example; structured discussion or peer review; independent reading and preparation.
Laboratory / seminar requirements: Backpropagation derivation. A tested implementation, benchmark, experiment report or technical design. Use only synthetic, public, or explicitly approved data; record procedures and limitations.
Security & governance checkpoint: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08; identify at least one relevant control, potential misuse or failure mode, and proportionate safeguard.
Assessment evidence: Submit a dated lab log or seminar brief with objective, methods, results, references, limitations and reflection.
Reading assignment: Relevant selections from the course reading list, assigned and verified by the instructor before the term.
Week 06 — CNNs, sequence models and inductive bias
Weekly objectives: (1) Explain the principal concepts, assumptions and limitations of cnns, sequence models and inductive bias. (2) Apply the week's methods in architecture comparison. (3) Engineer reproducible experiments, evaluate adversarial risks and define operational evidence. (4) Interpret evidence and communicate at least one limitation or trade-off.
Instructional plan: Instructor-led conceptual seminar; guided worked example; structured discussion or peer review; independent reading and preparation.
Laboratory / seminar requirements: Architecture comparison. A tested implementation, benchmark, experiment report or technical design. Use only synthetic, public, or explicitly approved data; record procedures and limitations.
Security & governance checkpoint: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08; identify at least one relevant control, potential misuse or failure mode, and proportionate safeguard.
Assessment evidence: Submit a dated lab log or seminar brief with objective, methods, results, references, limitations and reflection.
Reading assignment: Relevant selections from the course reading list, assigned and verified by the instructor before the term.
Week 07 — Attention and transformer mechanics
Weekly objectives: (1) Explain the principal concepts, assumptions and limitations of attention and transformer mechanics. (2) Apply the week's methods in attention implementation and scaling. (3) Engineer reproducible experiments, evaluate adversarial risks and define operational evidence. (4) Interpret evidence and communicate at least one limitation or trade-off.
Instructional plan: Instructor-led conceptual seminar; guided worked example; structured discussion or peer review; independent reading and preparation.
Laboratory / seminar requirements: Attention implementation and scaling. A tested implementation, benchmark, experiment report or technical design. Use only synthetic, public, or explicitly approved data; record procedures and limitations.
Security & governance checkpoint: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08; identify at least one relevant control, potential misuse or failure mode, and proportionate safeguard.
Assessment evidence: Submit a dated lab log or seminar brief with objective, methods, results, references, limitations and reflection.
Reading assignment: Relevant selections from the course reading list, assigned and verified by the instructor before the term.
Week 08 — Experimental rigor and midterm
Weekly objectives: (1) Explain the principal concepts, assumptions and limitations of experimental rigor and midterm. (2) Apply the week's methods in leakage audit, confidence intervals and exam. (3) Engineer reproducible experiments, evaluate adversarial risks and define operational evidence. (4) Interpret evidence and communicate at least one limitation or trade-off.
Instructional plan: Instructor-led conceptual seminar; guided worked example; structured discussion or peer review; independent reading and preparation.
Laboratory / seminar requirements: Leakage audit, confidence intervals and exam. A tested implementation, benchmark, experiment report or technical design. Use only synthetic, public, or explicitly approved data; record procedures and limitations.
Security & governance checkpoint: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08; identify at least one relevant control, potential misuse or failure mode, and proportionate safeguard.
Assessment evidence: Submit a dated lab log or seminar brief with objective, methods, results, references, limitations and reflection.
Reading assignment: Relevant selections from the course reading list, assigned and verified by the instructor before the term.
Week 09 — Foundation models and adaptation
Weekly objectives: (1) Explain the principal concepts, assumptions and limitations of foundation models and adaptation. (2) Apply the week's methods in prompting vs fine-tuning experiment. (3) Engineer reproducible experiments, evaluate adversarial risks and define operational evidence. (4) Interpret evidence and communicate at least one limitation or trade-off.
Instructional plan: Instructor-led conceptual seminar; guided worked example; structured discussion or peer review; independent reading and preparation.
Laboratory / seminar requirements: Prompting vs fine-tuning experiment. A tested implementation, benchmark, experiment report or technical design. Use only synthetic, public, or explicitly approved data; record procedures and limitations.
Security & governance checkpoint: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08; identify at least one relevant control, potential misuse or failure mode, and proportionate safeguard.
Assessment evidence: Submit a dated lab log or seminar brief with objective, methods, results, references, limitations and reflection.
Reading assignment: Relevant selections from the course reading list, assigned and verified by the instructor before the term.
Week 10 — Parameter-efficient fine-tuning
Weekly objectives: (1) Explain the principal concepts, assumptions and limitations of parameter-efficient fine-tuning. (2) Apply the week's methods in lora/adapter design review. (3) Engineer reproducible experiments, evaluate adversarial risks and define operational evidence. (4) Interpret evidence and communicate at least one limitation or trade-off.
Instructional plan: Instructor-led conceptual seminar; guided worked example; structured discussion or peer review; independent reading and preparation.
Laboratory / seminar requirements: LoRA/adapter design review. A tested implementation, benchmark, experiment report or technical design. Use only synthetic, public, or explicitly approved data; record procedures and limitations.
Security & governance checkpoint: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08; identify at least one relevant control, potential misuse or failure mode, and proportionate safeguard.
Assessment evidence: Submit a dated lab log or seminar brief with objective, methods, results, references, limitations and reflection.
Reading assignment: Relevant selections from the course reading list, assigned and verified by the instructor before the term.
Week 11 — RAG architecture and retrieval evaluation
Weekly objectives: (1) Explain the principal concepts, assumptions and limitations of rag architecture and retrieval evaluation. (2) Apply the week's methods in retriever/reranker/grounding pipeline. (3) Engineer reproducible experiments, evaluate adversarial risks and define operational evidence. (4) Interpret evidence and communicate at least one limitation or trade-off.
Instructional plan: Instructor-led conceptual seminar; guided worked example; structured discussion or peer review; independent reading and preparation.
Laboratory / seminar requirements: Retriever/reranker/grounding pipeline. A tested implementation, benchmark, experiment report or technical design. Use only synthetic, public, or explicitly approved data; record procedures and limitations.
Security & governance checkpoint: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08; identify at least one relevant control, potential misuse or failure mode, and proportionate safeguard.
Assessment evidence: Submit a dated lab log or seminar brief with objective, methods, results, references, limitations and reflection.
Reading assignment: Relevant selections from the course reading list, assigned and verified by the instructor before the term.
Week 12 — Explainability and causal limits
Weekly objectives: (1) Explain the principal concepts, assumptions and limitations of explainability and causal limits. (2) Apply the week's methods in attribution stability assessment. (3) Engineer reproducible experiments, evaluate adversarial risks and define operational evidence. (4) Interpret evidence and communicate at least one limitation or trade-off.
Instructional plan: Instructor-led conceptual seminar; guided worked example; structured discussion or peer review; independent reading and preparation.
Laboratory / seminar requirements: Attribution stability assessment. A tested implementation, benchmark, experiment report or technical design. Use only synthetic, public, or explicitly approved data; record procedures and limitations.
Security & governance checkpoint: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08; identify at least one relevant control, potential misuse or failure mode, and proportionate safeguard.
Assessment evidence: Submit a dated lab log or seminar brief with objective, methods, results, references, limitations and reflection.
Reading assignment: Relevant selections from the course reading list, assigned and verified by the instructor before the term.
Week 13 — Robustness, adversarial ML and AI security
Weekly objectives: (1) Explain the principal concepts, assumptions and limitations of robustness, adversarial ml and ai security. (2) Apply the week's methods in authorized attack-defense test. (3) Engineer reproducible experiments, evaluate adversarial risks and define operational evidence. (4) Interpret evidence and communicate at least one limitation or trade-off.
Instructional plan: Instructor-led conceptual seminar; guided worked example; structured discussion or peer review; independent reading and preparation.
Laboratory / seminar requirements: Authorized attack-defense test. A tested implementation, benchmark, experiment report or technical design. Use only synthetic, public, or explicitly approved data; record procedures and limitations.
Security & governance checkpoint: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08; identify at least one relevant control, potential misuse or failure mode, and proportionate safeguard.
Assessment evidence: Submit a dated lab log or seminar brief with objective, methods, results, references, limitations and reflection.
Reading assignment: Relevant selections from the course reading list, assigned and verified by the instructor before the term.
Week 14 — MLOps, drift, observability and governance gates
Weekly objectives: (1) Explain the principal concepts, assumptions and limitations of mlops, drift, observability and governance gates. (2) Apply the week's methods in monitoring and rollback design. (3) Engineer reproducible experiments, evaluate adversarial risks and define operational evidence. (4) Interpret evidence and communicate at least one limitation or trade-off.
Instructional plan: Instructor-led conceptual seminar; guided worked example; structured discussion or peer review; independent reading and preparation.
Laboratory / seminar requirements: Monitoring and rollback design. A tested implementation, benchmark, experiment report or technical design. Use only synthetic, public, or explicitly approved data; record procedures and limitations.
Security & governance checkpoint: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08; identify at least one relevant control, potential misuse or failure mode, and proportionate safeguard.
Assessment evidence: Submit a dated lab log or seminar brief with objective, methods, results, references, limitations and reflection.
Reading assignment: Relevant selections from the course reading list, assigned and verified by the instructor before the term.
Week 15 — Research replication and system defense
Weekly objectives: (1) Explain the principal concepts, assumptions and limitations of research replication and system defense. (2) Apply the week's methods in reproducible artifacts and oral defense. (3) Engineer reproducible experiments, evaluate adversarial risks and define operational evidence. (4) Interpret evidence and communicate at least one limitation or trade-off.
Instructional plan: Instructor-led conceptual seminar; guided worked example; structured discussion or peer review; independent reading and preparation.
Laboratory / seminar requirements: Reproducible artifacts and oral defense. A tested implementation, benchmark, experiment report or technical design. Use only synthetic, public, or explicitly approved data; record procedures and limitations.
Security & governance checkpoint: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08; identify at least one relevant control, potential misuse or failure mode, and proportionate safeguard.
Assessment evidence: Submit a dated lab log or seminar brief with objective, methods, results, references, limitations and reflection.
Reading assignment: Relevant selections from the course reading list, assigned and verified by the instructor before the term.
Indicative reading list
- Bishop, Pattern Recognition and Machine Learning or Deep Learning: Foundations and Concepts, selected chapters.
- Goodfellow, Bengio & Courville, Deep Learning, selected chapters.
- Vaswani et al. (2017), “Attention Is All You Need.”
- Hu et al. (2022), “LoRA: Low-Rank Adaptation of Large Language Models.”
- Lewis et al. (2020), “Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks.”
- Selected contemporary reproducibility, robustness and evaluation papers, verified by instructor.
Capstone and evaluation evidence
The final submission must state the problem and constraints, method and baseline, data/source provenance, reproducible procedures, quantitative or qualitative evaluation, limitations, ethical and security implications, and evidence supporting conclusions. Include a short individual oral defense. Instructor will publish detailed rubrics and resource/accessibility alternatives before enrollment.
Quality assurance before academic adoption
Faculty peer review; verification of reading accessibility and rights; learning-outcome-to-assessment mapping; workload and credit-hour calibration; disciplinary prerequisites; inclusive pedagogy review; and alignment with the adopting institution's academic governance process.
Curriculum alignment and instructor delivery
This course is part of the proposed IAIGCRM Secure & Governed AI Development Curriculum. It is designed for potential delivery by IAIGCRM faculty or the Institute's founder, subject to final staffing, scheduling, enrollment, institutional agreements and resource availability. No enrollment, credits or accreditation are currently offered by this draft. The graduate integrative capstone is AI-GR551. Security baseline: SEC-02, SEC-03, SEC-04, SEC-06, SEC-08.
Weekly teaching preparation and laboratory quality gate
For each class, prepare slides, instructor demonstration, learner instructions, environment setup, starter files, a marking rubric, accessibility alternatives, and a verified reading packet. Labs must include a learning objective, prerequisites, data provenance, stepwise tasks, tests, expected outputs, failure cases, evidence submission and cleanup procedure. Students may not attack real systems or use non-consensual personal data.