Assignments Management
CS704 Applied Agentic AI • Dr. Sarah Bennett • Spring 2027
Total6All curriculum tasks
Open3Accepting work
Grading1Action required
Completed1Grades published
Draft1Unpublished specs
CS704Grading in ProgressOral Viva Required (15% weight)
A2: Autonomous Research Agent with Grounded SEC 10-K RAG
Students build a multi-turn agent capable of complex SEC financial document retrieval, multi-hop reasoning, and indirect prompt injection resistance.
42/ 48 submitted
Cohort Avg: 81.2 / 10032 finalised • 10 awaiting reviewRubric & Marks10 Criteria • 100 Marks
Deadline & CutoffCloses in 2 days (Grace Period)
Sandbox Execution ProfilePython 3.11 • LangGraph • gVisor Container • Egress Blocked
CS704Completed & Released
A1: Deterministic Tool Calling & Structured Schema Validation
Implement JSON-schema bounded tool invocation with error recovery loops and deterministic execution guards.
48/ 48 submitted
Cohort Avg: 86.2 / 100Rubric & Marks6 Criteria • 100 Marks
Deadline & CutoffClosed 12 days ago
Sandbox Execution ProfilePython 3.11 • Pydantic v2 • Isolated Docker • Mock Services
CS704Open for SubmissionsOral Viva Required (20% weight)
A3: Multi-Agent Supervisor Swarm with Adversarial Negotiation
Design a hierarchical agent coordinator orchestrating specialist worker agents with contract validation.
4/ 48 submitted
Rubric & Marks8 Criteria • 100 Marks
Deadline & CutoffCloses in 23 days
Sandbox Execution ProfilePython 3.12 • Autogen / CrewAI • gVisor VM • Rate Limited Egress
CS704Open for Submissions
A4: Long-Horizon Memory & Self-Reflective Planning
Build an episodic vector memory layer with reflection passes and pruning heuristics.
0/ 48 submitted
Rubric & Marks7 Criteria • 100 Marks
Deadline & CutoffCloses in 44 days
Sandbox Execution ProfilePython 3.11 • Qdrant local • Sandboxed gVisor • 2GB RAM
CS704Open for SubmissionsOral Viva Required (25% weight)
A5: LLM Red-Teaming & Jailbreak Defense Benchmark
Evaluate defense mechanisms against automated prompt injection and privilege escalation attacks.
0/ 48 submitted
Rubric & Marks9 Criteria • 100 Marks
Deadline & CutoffCloses in 65 days
Sandbox Execution ProfilePython 3.11 • PyRIT / Garak Harness • Strict Network Quarantine
CS704Draft SpecificationOral Viva Required (30% weight)
A6: Final Capstone: Autonomous Financial Analyst System
Comprehensive end-to-end multi-modal agent deployed with production-grade monitoring and guardrails.
0/ 48 submitted
Rubric & Marks12 Criteria • 100 Marks
Deadline & CutoffDraft • Not visible to students
Sandbox Execution ProfileMulti-container orchestration • Kubernetes microVM • Monitored egress