Sylvain Kalache

AI-assisted incident response can lower MTTR while leaving engineers less prepared for the complex incidents automation cannot solve.

AI can generate code faster than humans can review it.

Sylvain Kalache went looking for the new ways AI breaks software.

Your AI agents are lost — why context engineering and knowledge graphs give agents the structure they need for retrieval, memory, planning, reasoning, and learning from business outcomes.

AI vs.

Speaking on the Incident Fest Main Stage about staying reliable as AI-assisted development increases the rate at which software is created.

Why Twingate stopped shipping features to rebuild reliability — from active-active multi-region architecture and smaller blast radiuses to preserving human ownership as agentic coding accelerates delivery.

Meta-hosted summit in Menlo Park focused on applying AI, LLMs, and agents to software incidents and response.

How a chain of reasonable-sounding decisions led an AI coding agent to run terraform destroy against a live production database — and the guardrails that separate moving fast from losing everything.

Engineering leaders discuss when AI coding assistants accelerate development and when they introduce security, quality, and architectural risk.

At AI DevSummit New York: how AI-assisted coding is reshaping software delivery — the productivity gains, the surge in incident rates, and what reliability teams need to do about it.

Every pilot trains for engine failure, but most engineers face their first major incident unprepared — why communication, coordination, and decision-making need deliberate simulation practice.

LLM observability — how monitoring principles from MLOps apply to large language model applications, the gaps that went overlooked for nine years, and best practices for observing AI systems.

An MCP-native arena that pits AI models against each other in real-time Doom matches to benchmark spatial planning, adaptation, speed, accuracy, and cost.

Why traditional burnout detection fails in engineering teams.

The golden hour — why the first 15 minutes of an incident decide everything, from mitigation strategy to customer trust, and why human psychology matters more than tools.

From 600 to 6,000 — an 18-month transformation replacing legacy tooling with a purpose-built incident management platform, scaling operational responsibility from 600 SREs to 6,000 engineers across the org.

AI didn't change the game, it just exposed your bottlenecks — how platform engineering and SRE teams solve identical human problems through influence rather than authority, and why AI amplifies existing bottlenecks instead of transforming operations.

Reliability Rebels Guest appearance to discuss reliability engineering, AI, and modern incident response.

Fear, identity, and flaky tests — why SRE resistance to AI agents stems from identity and control concerns rather than the technology itself, and practical strategies for adopting AI-driven reliability tools starting with low-risk tasks.

A candid conversation on software leadership, reliability engineering, and the evolving role of AI in modern engineering teams.

From faster detection to smarter debugging — where AI delivers practical value in incident response, its limitations in production environments, and how teams can integrate AI while maintaining reliability standards.

The incident you never had — deterministic simulation testing, why conventional testing misses bugs that cause real outages, and how simulation-based approaches improve software reliability.

LLM-assisted development introduces hard-to-trace bugs, AI-generated tests that mirror flawed logic, and hallucinated dependencies.

The science behind responder overload and introducing On-Call Health, an open-source tool for detecting incident responder fatigue.

Burnout doesn't ask permission — the physiology of burnout from an incident responder's perspective, warning signs, recovery, and the systemic changes the industry must make.

Code is cheap, reliability isn't — owning production in the AI era, the hidden complexity of SRE work, and why human ownership remains essential.

An evening of AI demos and networking in Toronto — presenting Rootly AI alongside teams from Sentry, Red Brick Labs, Convex, and DigitalOcean.

A deep dive into the Kubernetes support lifecycle and what it means for the massive number of clusters running end-of-life versions.

Democratizing reliability — empowering non-engineers with operational power, incident management as a muscle, and AI-powered postmortems.

99%+ accuracy on a moving target — model deprecation, reliability with LLMs, and treating prompts as architectural components.

Context switching kills developer productivity.

The reality of GenAI in production — why organizational culture is the biggest blocker, scaling non-deterministic LLM systems, and what separates AI winners from experimenters.

A conversation about developer advocacy, AI evangelism, and building reliable systems at scale.

Developer awareness, content strategy, and the human side of reliability engineering.

AI-generated code is outpacing traditional runbooks.

Release lead Vyom Yadav unpacks 58+ enhancements: DRA GA for GPU workloads, mutating admission policies, and major API server performance gains.

AI-assisted coding is exploding — but acceleration doesn't mean reliability.

Discussion at SREcon EMEA 2025 on how AI is transforming the software development lifecycle — CI/CD pipelines, deployments, scaling, monitoring, incident management, reliability tooling, and emerging disciplines like LLMOps.

Panel exploring AI-driven automation and observability with leaders from NVIDIA, OpenAI, Baseten, Replit, and Weights & Biases on scaling operations and reducing MTTR.

The end of good code, AI throughput, and what reliability means at CI/CD scale.

Panel at the AWS GenAI Loft in San Francisco on MCPs, incident automation, observability, and generative AI tooling for reliability.

Three emerging patterns from PlatformCon that signal how AI is reshaping internal developer platforms and platform team workflows.

Leading engineering at scale in e-commerce aggregation and the operational challenges that come with it.

Panel at Google HQ with 300+ attendees exploring real-world Gemini models, reinforcement learning, next-gen agent systems, and AI reliability.

Live demos at Okta HQ in San Francisco — presenting Rootly AI alongside cybersecurity startups tackling identity, endpoint security, and threat response.

Exploring whether AI-assisted coding tools will democratize software development or create new categories of hard-to-debug production issues.

Building and operating core banking infrastructure with a focus on uptime and regulatory compliance.

Interview with Alan Shimel at PlatformCon NYC 2025 on how AI applies to incident management and reliability engineering — triage, root cause analysis, and why AI enhances rather than replaces engineers.

Invite-only roundtable on filtering alert noise, enriching alerts with automated ownership context, and streamlining bug validation workflows.