Rishabh Banga
Owner, RBX Labs
Toronto, Canada
Actions
Rishabh Banga is an AI Product Leader focused on building production-grade AI systems across marketplaces, trust platforms, and ML-driven discovery. With 7+ years of experience, he has led 0→1 AI products generating $1M+ ARR and improving activation (20→31%) through ranking, experimentation, and governance systems.
Driven by a long-standing interest in bridging hardware and software - From drone logistics to AI platforms - Rishabh focuses on turning cutting-edge AI into reliable, scalable systems that work beyond demos.
Links
Area of Expertise
Topics
When AI Fails Silently: How Government Teams Detect Risk and Build Trust
As government organizations explore AI for search, service delivery, operations, and decision support, one of the biggest challenges is that AI systems often do not fail visibly. They fail silently through hallucinations, weak retrieval, inconsistent reasoning, unsafe outputs, and overconfident responses. These issues are difficult to catch with policy alone and can erode public trust quickly.
This session will focus on practical, vendor-neutral approaches that government technology leaders can use to evaluate and govern AI systems more effectively. It will cover common failure modes in modern AI systems, why traditional testing and static guardrails are often insufficient, and how teams can introduce trust signals, verification layers, human review, and monitoring practices into real-world workflows.
Attendees will leave with a clearer framework for thinking about AI risk in operational settings, along with practical ideas for moving from experimentation to more trustworthy implementation across public-sector environments.
From AI Pilots to Production: Detecting Silent Failures Before Users Do
Enterprise AI systems often do not fail in obvious ways. They continue to run, APIs stay healthy, and applications remain available — yet outputs can still be wrong, weakly grounded, inconsistent, or unsafe. This creates a new operational challenge for teams deploying AI in real-world products and workflows.
This session introduces a practical framework for detecting those silent failures before they reach users. It explores how trust can be treated as a runtime capability through signals such as grounding quality, contextual risk, and output consistency, rather than relying only on static guardrails or offline evaluation.
Designed for teams working on intelligent apps, Copilot-style experiences, and broader enterprise AI adoption, the session will show how to think about trust, governance, and observability as part of production architecture. Attendees will leave with a clearer model for moving from AI pilots to more reliable, production-ready systems.
Designing Trust Layers for AI: Scoring, Moderation & Governance in Real Time
Most AI systems today are optimized for accuracy, latency, and cost—but fail where it matters most: trust.
In production, AI rarely crashes. Instead, it fails silently—through hallucinations, unsafe outputs, biased decisions, and degraded user experiences. Traditional approaches like prompt engineering, offline evaluation, and static guardrails are not sufficient to detect or prevent these failures in real time.
This talk introduces a new architectural primitive: the Trust Layer.
We’ll walk through how to design and implement real-time trust scoring systems (0–100) that evaluate AI outputs before they reach users. By combining signals across model confidence, retrieval quality, behavioral patterns, and contextual risk, teams can move from reactive debugging to proactive reliability.
Through real-world examples, we’ll cover:
Why current guardrails fail in production environments
Designing multi-signal trust scoring systems
Integrating trust layers into RAG pipelines, agent workflows, and ranking systems
Building observability to detect silent failures early
Attendees will leave with a practical blueprint to build more reliable, production-grade AI systems.
Catching Silent Failures in MCP Workflows: Trust, Verification, and Observability for Agents
MCP gives developers a standardized way to connect models, tools, and applications, but successful connectivity does not guarantee reliable behavior. In production, many failures are silent: a tool call can succeed, the protocol can behave as expected, and the system can remain available, while the agent still uses the wrong context, takes the wrong action, or produces an unsafe result.
This session explores how to build MCP workflows with stronger runtime safeguards. It will cover practical approaches for introducing trust signals, verification checks, and observability into MCP-based systems so teams can better understand agent behavior and detect failure modes before they cascade across multi-step workflows.
The goal is to move the conversation beyond connectivity alone and toward operational maturity. Attendees will leave with a clearer framework for designing MCP systems that are not only interoperable, but also more reliable, transparent, and production-ready.
Designing Trust Layers for AI: Scoring, Moderation & Governance in Real Time
Most AI systems today are optimized for accuracy, latency, and cost—but fail where it matters most: trust.
In production, AI rarely crashes. Instead, it fails silently—through hallucinations, unsafe outputs, biased decisions, and degraded user experiences. Traditional approaches like prompt engineering, offline evaluation, and static guardrails are not sufficient to detect or prevent these failures in real time.
This talk introduces a new architectural primitive: the Trust Layer.
We’ll walk through how to design and implement real-time trust scoring systems (0–100) that evaluate AI outputs before they reach users. By combining signals across model confidence, retrieval quality, behavioral patterns, and contextual risk, teams can move from reactive debugging to proactive reliability.
Through real-world examples, we’ll cover:
Why current guardrails fail in production environments
Designing multi-signal trust scoring systems
Integrating trust layers into RAG pipelines, agent workflows, and ranking systems
Building observability to detect silent failures early
Attendees will leave with a practical blueprint to build more reliable, production-grade AI systems.
Building AI Trust in Production: A Practical Framework
Most AI systems perform well in controlled demos but fail to scale in production. The gap is not just model quality, it is trust.
This session introduces a practical framework for building trust into AI systems so they can move from promising pilots to reliable, production-ready experiences. We will break down what “trust” actually means in deployed systems, including confidence scoring, risk signals, moderation layers, and operational visibility.
Using a real product journey as a case lens, this talk explores how trust can be designed as a system layer - Connecting model outputs with user safety, governance, and decision-making. We will cover how to instrument trust signals, define escalation paths, and integrate monitoring that goes beyond traditional observability.
Attendees will leave with a clear, actionable framework to design, evaluate, and operationalize trust in their own AI systems - whether they are building consumer products, enterprise tools, or platform-level infrastructure.
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top