Exclusive 40% Discount: Harness Engineering – Build an AI Data Analyst You Can Actually Trust
Special 40% Discount for the IBridgePy Community
Save 40% using promo code HUI40 or enroll directly via our partner link below.
When AI agents hallucinate revenue explanations, crash under overwhelming tool outputs, or execute unauthorized database commands, the problem isn't the model—it's the harness surrounding it.
In this 4-hour, hands-on workshop led by international bestselling author Bruno Gonçalves, PhD, you will build an autonomous AI data analyst in pure Python and engineer the battle-tested harness around it: connecting permissioned MCP (Model Context Protocol) tools, enforcing strict token budgets, implementing multi-layer guardrails, and verifying mathematical numbers against re-runnable database queries.
IBridgePy has partnered with Packt to bring our algorithmic trading community an exclusive 40% discount with coupon code HUI40.
📅 Workshop Snapshot
Saturday, October 31, 2026
9:30 AM – 1:30 PM EDT (1:30 – 5:30 PM UTC)
4 Hours • Hands-on Live Interactive Workshop
Bruno Gonçalves, PhD (Founder Data For Science)
Packt - ML & AI Engineering
40% Off with Code HUI40 ($89.99 vs $149.99)
Ready to Build Trustworthy AI Agents with MCP?
Build without framework bloat. Learn how to engineer robust guardrails, token budgets, and verification checks. Save 40% with code HUI40.
Enroll with 40% Discount (Code: HUI40) →Includes live workshop access, complete reusable Python harness, fault-injected test suites, and recording.
How Harness Engineering Empowers IBridgePy Quant Traders
Automated trading generates thousands of log entries, execution fills, slippage reports, and market data queries. Many quantitative traders want an intelligent AI assistant to monitor their live trading, audit performance, and query databases—but they cannot risk an unconstrained agent executing unintended database mutations or making decisions based on fabricated numbers.
Here is how this workshop directly benefits IBridgePy traders:
- Safe Auditing of IBridgePy Trade Logs: Use permissioned MCP tools to connect your AI analyst to DuckDB or SQLite databases storing historical IBridgePy fills and order execution histories without giving the agent raw write access.
- Rigorous Number Verification: Financial data requires exact accuracy. You will learn to force the agent to anchor every statistic (e.g. Sharpe ratio, max drawdown, win rate) to a re-executable SQL query rather than relying on LLM arithmetic.
- Fault-Tolerant Supervised Execution: When market data connections glitch or queries time out, harness supervisors handle automatic retries, execution budgets, and structured traces so your analysis pipeline remains resilient.
- Framework-Free Control: Because you build from first principles in clean Python without heavyweight wrapper frameworks, you have 100% control over how your agent harness interacts with your trading environment.
What You Will Master:
- Permissioned MCP Tool Integration: Discover and execute tools from a third-party DuckDB MCP server, validate parameter schemas, and enforce safety policies before actions touch the server.
- Strict Context & Token Budget Management: Implement compact result summaries, expandable data handles, conversation compaction, and tiered memory so your agent never overflows token limits.
- Multi-Layer Guardrails: Separate instructions from retrieved data, defend against indirect prompt injection, and inspect SQL statements to block unauthorized writes and network requests.
- Numerical & Claim Verification: Anchor every numerical insight in an AI report to a re-runnable query, checking units, periods, and bounds.
- Diagnosable Failure Handling: Implement timeouts, circuit breakers, execution budgets, and structured traces tested against a fault-injected validation suite.
Meet Your Instructor: Bruno Gonçalves, PhD
Bruno Gonçalves is the Founder of Data For Science, an author, international keynote speaker, and consultant specializing in Generative AI, Machine Learning, and Complex Systems.
With a PhD in Physics of Complex Systems, Bruno was previously a tenured faculty member at Aix-Marseille Université and a Data Science Fellow at NYU’s Center for Data Science. He has trained thousands of practitioners through O’Reilly, Pearson, and live masterclasses. His open-source harness notebooks explore production agent architectures from first principles.
Included with Your Registration:
- Working AI Data Analyst & Reusable Python Harness: 4 production-grade modules covering tools, memory, guardrails, and supervision.
- Fault-Injected Test Suite: Reusable CI/CD validation tests to stress-test your own agent harnesses.
- Full HD Session Recording: Lifetime access to re-watch every walkthrough and code demonstration.
- Complete Slide Deck & Checkpoint Notebooks: Detailed presentation materials and clean stage checkpoints.
- Official Certificate of Participation: Verified credential issued by Packt recognizing your training in harness engineering.
Frequently Asked Questions
Click any enrollment button on this page. Promo code HUI40 will be automatically applied at Luma checkout, saving 40% on admission ($89.99 vs $149.99).
No. The workshop runs entirely locally on your laptop using a mock model, a real DuckDB MCP server, and a local database. No GPU or paid cloud account is needed.
Yes. All registered attendees receive full unlimited access to the recording and all accompanying code checkpoints.
Build an AI Analyst You Can Truly Rely On
Take control of your agent's tools, memory, security, and outputs. Use coupon code HUI40 to claim your 40% discount now.
Register on Luma with 40% Off →Discount applied automatically • 100% Online • Python Harness & Recordings Included