Loading...
AI Innovation without Limits
We're shaping the future of cybersecurity with AI. From the products we ship to automated workflows, AI powers everything we do. If you're ready to solve problems that don't have answers yet, your curiosity has a home here.
What Sets Us Apart
AI-Powered Velocity
While other companies figure out how to use AI, we're building with it as our foundation. We're not just coding faster - we're building what wasn't possible before.
Real Ownership
You're not just following specs here. You'll challenge assumptions and solve cybersecurity problems that matter. Autonomy comes standard and is backed by trust.
Work That Matters
Everything you build protects real people from threats that could change their lives. You'll stop breaches before they happen and protect what's important.
No Playbooks
We're fighting threats using AI and legacy tools can’t keep up. Our platform evolves as fast as attackers because threats don’t wait for traditional solutions.
Our Approach
We Are Curious
Experimentation is part of the process, not something we do for fun. You’re trusted to explore bold ideas, prototype quickly and ship solutions fast.
We Think AI-First
ML is critical to what we build and how we work. It's embedded in how we plan, code, and collaborate, helping us tackle challenges others can't.
We Work Flexibly
Our culture is flexible by design, built on async clarity, shared knowledge, and smart systems that keep teams aligned across continents and time zones.
Growth Through Ownership
Growth doesn't follow a script here. You'll grow by taking initiative, learning out loud, and designing solutions that don't exist yet. And when you push great ideas forward, we give you the tools, time and support to make them happen.
How We Actually Use AI
AI transforms how engineering happens here. It's embedded in every layer of how we experiment, plan, design, and build.
Claude Code for On-Call Ops
When a service breaks in the middle of the night, engineers often wake up to a vague page that simply says “Service is broken.” To address this, Shrivu Shankar developed Claude Code for On-Call Ops, a tool that automatically connects to CloudWatch and Prometheus, executes the correct queries, and generates a clear diagnostic report. The result: engineers can respond quickly without having to juggle multiple dashboards or query syntaxes.
Nora Incident Responder
The Nora Incident Responder is an AI assistant created by Rishi Kavikondala that automates on-call triage in Slack. It analyzes PagerDuty alerts, gathers metrics, logs, and runbook details, and posts a concise summary with next steps, allowing engineers to focus on fixing issues instead of manual investigation. NOTE: Demo visuals use either blurred real data or synthetic placeholders to protect customer privacy.
Application Development Platform
Building new services used to take days, slowed by missing documentation and inconsistent environments. Matthew Westberg built the Application Development Platform, an AI-driven CLI that embeds Abnormal’s engineering patterns to scaffold Go and Python gRPC services, connect Postgres and S3 storage, and spin up Streamlit front ends.
Jira Orchestration for Initiatives
Product managers often spend hours writing PRDs and breaking initiatives into epics and tasks. To fix this, Tushar Amrit built a Jira Orchestrator powered by Nora. It generates structured PRDs, proposes epics, and proactively suggests next steps, saving time and letting humans focus on review instead of repetitive creation. NOTE: Demo visuals use either blurred real data or synthetic placeholders to protect customer privacy.
Jira Orchestration for Initiatives - Establishing the Foundation (1)
Product managers often spend hours writing PRDs and breaking initiatives into epics and tasks. To fix this, Tushar built a Jira Orchestrator powered by Nora.
Nora Tech Plan
Traditional tech design docs (TDDs) were built for humans to read, not for AI to act on. Shrivu Shankar reimagines that process with Nora Tech Plan, generating documents that both capture architectural intent and serve as direct inputs for AI code generation.
Claude Code Feedback Loop
AI agents already generate a large share of Abnormal’s code, but when they hit errors in CI environments, the problems often go unnoticed. Shrivu Shankar built the Claude Code Feedback Loop, a system that pipes agent logs into an LLM, extracts failure patterns, and feeds fixes back into development.
Experimentation Engine
By combining data science and AI, Amal Bhatnagar and the MEP team reduced false positives in Abnormal’s Outbound Email Security product by 85% and cut feature validation time from weeks to days, accelerating innovation for zero-to-one detection systems.
Nora Deep Research
Nora Deep Research V3 connects unstructured data across Gong, Salesforce, Jira, and more to help teams instantly uncover insights, size product impact, and even map customer pain directly to code changes.
Product Feedback Pulse
Customer feedback fuels innovation, but it’s often scattered across systems and difficult to synthesize. To fix that, Ellie Kloberdanz built an AI agent that consolidates customer feedback by product and timeframe, clusters it into themes, and delivers automated summaries straight to your inbox.
AI Optimization of AWS Compute Speed
Optimizing backend services is complex and time-consuming. To simplify it, Daniel Ferstay built an AI-enabled system that offloads the hardest parts of performance tuning to Claude. The workflow identifies inefficiencies in heap dumps and dashboards, recommends code changes, and right-sizes container resources, all while teaching engineers best practices along the way.
Nora Monitoring Dashboards
When Nora slows down or hits an error, teams need to know why. Tushar Amrit built a comprehensive monitoring solution using Grafana and Databricks, powered by Nora Plan and Nora PR, that tracks Nora’s health, usage, and adoption trends without compromising user privacy.
Posture Card GPT
Rohan Talathi, a Technical Program Manager on Abnormal’s R&D Engine team, built Posture Card GPT to accelerate the creation of posture content for Abnormal’s Security Posture Management (SPM) division. His prototype transforms what was once a multi-week, engineering-heavy task into a drag-and-drop workflow powered by AI.
Salesforce Nora Evidence
Tushar Amrit, a software engineer on the Gen AI team, built Nora Salesforce Evidence, a sub-agent that links Nora directly to Salesforce. The goal was simple: help product managers generate clean, complete customer reports in seconds, not hours. NOTE: Demo visuals use either blurred real data or synthetic placeholders to protect customer privacy.
Nora Feedback Analyzer
Tushar Amrit enhanced Nora’s feedback process by turning basic emoji reactions into actionable insights. The new system extracts feedback from user conversations, generates weekly performance reports, highlights missing documentation, and even suggests how to refine Nora’s prompts, enabling AI to improve itself over time.
Post Incident Ops
Ivan Penev, a senior software engineer, built Post Incident Ops. It's an agent that gathers incident telemetry and drafts Abnormal’s post-mortem template from a raw data dump. The result is faster, complete timelines and clearer action items, reducing fatigue for on-call engineers and accelerating prevention work. NOTE: Demo visuals use either blurred real data or synthetic placeholders to protect customer privacy.
Nora PR 'V2'
As Abnormal accelerates toward an AI-initiated engineering culture, tools must support fast, reliable, and scalable code generation. Shrivu Shankar redesigned NoraPR from the ground up, fixing naming confusion, speeding up CI performance, and introducing a local mode that encourages AI-driven development while still providing guardrails.
Nora PR Analyzer
As AI writes more of Abnormal’s code, context quality becomes the limiting factor. Shrivu Shankar built NoraPR Analyzer to close that gap by analyzing how AI-written code differs from human-written code and automatically generating the documentation needed to make AI better next time.
Code to Content - Agent Trio Pipeline (4)
Brandon Qin’s latest Code to Content iteration adds a third agent for visuals and a feedback system that learns from edits, enabling faster, higher-quality AI-generated websites.
RFE Prioritizer
This feature, by Brandon Qin (AI Native Product Manager), helps product managers triage thousands of Jira enhancement requests by enriching tickets with Salesforce ARR and Gong context, then emailing ranked recommendations with rationale. The V1 reduces manual backlog grooming and surfaces renewal risk before quarterly escalations.
Nora PCCMR
Software engineer Ivan Penev built a command-line agent to reduce the toil of production change reviews. It gathers fragmented context, asks clarifying questions, generates detailed JIRA-ready change review tickets, and raises consistency so risky changes ship with safer execution, monitoring, and rollback plans.
Sigma CLI
Sigma is widely used across Abnormal, but its UI-heavy workflows made even small analytics changes slow and painful. Shrivu Shankar built Sigma CLI to bypass the interface entirely, connecting Claude Code directly to Sigma’s underlying data and unlocking fast, AI-driven analytics.
Modal Dev Boxes
Modal Dev Boxes gives Abnormal engineers instant, secure agent workspaces in the browser. Shahzil Sheikh shows how AI agents can use real tools, access internal resources securely, and reduce setup time from minutes or hours to seconds, enabling faster shipping today.
Constitutional Markdown
Shrivu Shankar built Constitutional Markdown to help Abnormal operate as an AI-initiated org, where agents do the first pass on design and implementation. The feature packages Abnormal’s architecture, security, and legal intent into a single, agent-friendly interface.
Onboarding Scheduler
As Abnormal’s R&D onboarding became more hands-on and AI-first, scheduling enablement sessions by hand no longer scaled. Chanel Green built a self-serve onboarding scheduler using Claude Code and Google Apps Script to automate session scheduling directly from onboarding spreadsheets.
Enterprise Context Layer
As product velocity increases, documentation quickly falls out of sync with reality. Andy Chen and Sara Faradji are building a Claude Code–powered documentation layer that automatically updates knowledge bases, reconciles conflicting sources, and ships customer-facing updates within minutes of code changes.
Enterprise Context Layer - Automating Documentation (2)
As product velocity increases, documentation quickly falls out of sync with reality. Andy Chen and Sara Faradji are building a Claude Code–powered documentation layer that automatically updates knowledge bases, reconciles conflicting sources, and ships customer-facing updates within minutes of code changes.
Enterprise Context Layer - Building the Foundations (1)
Andy Chen enhanced Enterprise Context Layer to convert real customer questions into governed FAQs. It routes unanswered bot queries through classification, AI-generated drafts, and human approval, so customers get faster answers without exposing sensitive internal documentation or workflows.
Prompt to Demo Portal
When you’re iterating on Abnormal’s customer-facing portal UI, review friction is real. A PR might contain the “right” code, but most reviewers still can’t easily answer the simplest question: what does it look like? The Prompt to Demo Portal is Shrivu’s quick solution to collapse the process into a single Slack prompt that generates both the change and a demo video.
Ops-Dash (SaaS Vendor Status & Security Dashboard)
Tyler Takaro and Elvin Rivera built Ops-Dash, a SaaS vendor status and security dashboard for internal IT and ops teams. They took an early prototype and pushed it into a production-ready, self-hosted web app on AWS, tuned for Abnormal’s workflows.
CLI Tools for Agents
Ivan Penev shipped CLI Tools for Agents so Abnormal’s coding agents, Claude and Nora, can reliably retrieve operational and analytics context during real investigations. The goal was simple: fewer stalled sessions, more answers engineers can act on.
AI Marketing Campaign - From Brief to Banner (1)
Andres Lam built AI Marketing Campaign to reduce manual marketing handoffs by generating an AI-native brief and LinkedIn banner ad variants enriched with Salesforce. It streamlines targeting, messaging, and creative inputs, enabling teams to move faster from planning to publishing.
Customer Pulse
Hala Abualtayeb built a unified planning portal to solve a common challenge across product and engineering: understanding what customers are actually struggling with before it turns into escalations. By combining multiple data sources and using AI to cluster themes, the system enables more proactive planning and smarter prioritization.
Nora Research: Customer Voice-to-PR
As part of the broader “customer voice to PR” initiative, Nora Research is designed to transform customer feedback into production-ready fixes. But early pilots revealed key friction points, from overly verbose reports to unnecessary code generation. This latest iteration focuses on making the system faster, clearer, and more actionable.
Meet the Team
We operate with deep technical expertise, shared accountability, and curiosity to solve what others can't. We scale fast, challenge assumptions with data, and apply machine learning to solve complex cybersecurity problems.
Meet Abnormal’s Leadership Team
How We Deliver:
1
We treat AI as a true partner, not just an accelerator
2
We aren’t afraid of big questions and bold experimentation
3
We take full ownership of problems and outcomes
4
We aren’t afraid of ambiguity and find clarity in iteration
5
We care as much about quality as we do about velocity
Your Future, Engineered by AI
We’re looking for those who are ready to push what’s possible with AI to reimagine what modern engineering can look like. If that challenge excites you, we'd like to get to know you.