Software Engineer, AI Agent

Most AI agents are wrappers around a model. You're here to build something that actually thinks. If you want to work at the frontier of agentic AI — not prototype it, but ship it at scale — read on.

Sendbird is on a mission to build the AI workforce of tomorrow. For over a decade, we built the infrastructure behind conversations—chat, voice, video, messaging APIs—and became the #1 CPaaS platform for in-app communications. 4,000+ brands trust us. 7 billion messages flow through our platform every month. 300 million monthly active users.

We powered conversations for DoorDash, Match Group, Noom, Yahoo Sports, Rakuten, and thousands of others. We were good at what we did. Really good.

We also saw it early: AI would fundamentally reshape how businesses talk to customers. The infrastructure we'd spent a decade building would become commoditized. The value would move up the stack—into intelligence, into experience, into outcomes.

We had a choice: protect what we built, or reinvent ourselves.

We chose reinvention.

In December 2024, we made the full strategic pivot to AI-first customer experience. By February 2025, we'd launched our AI agent for enterprise CX—built on a decade of conversation data, now with intelligence on top. And in November 2025, we rebranded to Delight.ai.

The name says it all. AI's real promise isn't efficiency or cost savings. It's giving customers back something they lost—the feeling of being truly understood and cared for. Not satisfied. Delighted.

Delight.ai is the AI concierge for customer experience. Most AI agents forget you the moment the conversation ends. Ours doesn't. Delight.ai builds memory over time, learns preferences, and connects context across every channel—chat, SMS, email, voice, WhatsApp—without losing the thread. We're building AI that makes customers feel understood, seen, and remembered.

Most of what's being shipped in the AI agent space is shallow. Thin wrappers, brittle pipelines, demos that don't hold up in production. We're building differently. Sendbird's AI Agent platform is engineered for scale, reliability, and real intelligence, and the architectural decisions made now will define how millions of customer conversations are handled for years to come.

This isn't a role where you maintain someone else's system. The core agent platform, orchestration, RAG pipelines, evaluation, and observability are being shaped right now. You'll have significant influence over how it gets built, using agentic CLIs, LLM APIs, and automated pipelines to move faster and build smarter than a traditional engineering team ever could.

You'll own end-to-end delivery of core AI agent capabilities, from architecture to production, driving excellence across orchestration, prompt engineering, and platform scalability. You'll mentor the engineers around you and set the standard for AI-native development. This role is built for someone who moves fast, thinks in systems, and uses AI to ship at a pace that surprises people.

• You read AI research papers on your own time and have opinions about what's actually useful vs. overhyped.
• You've owned a system that broke in production, made it better, and still think about why it broke.
• You default to first principles on new architectural problems instead of reaching for the nearest framework.
• You use agentic tools like Claude Code or Codex as a core part of your dev workflow, not as a novelty.
• You get uncomfortable when a system is "working" but not well-instrumented. Observability isn't an afterthought for you.
• You enjoy mentoring junior engineers not because it's expected, but because you genuinely care about how they grow.
• You've shipped LLM-powered features into production and have strong views on where prompt engineering ends and architecture begins.
• You've led technical planning in a cross-functional setting and can hold your own in a room with PMs, designers, and engineers at once.

• 3+ years of software engineering experience with direct ownership of AI/ML or backend systems in production.
• Deep Python expertise with a proven track record of building and scaling reliable, performant systems.
• Hands-on experience with LLM APIs and agentic frameworks such as LangChain or LangGraph.
• Strong system design and architecture skills, including distributed systems and cloud-native deployments on AWS or GCP.
• Strong English proficiency. You can communicate complex technical ideas clearly in writing and in conversation.

• Design and implement core agent features including advanced RAG pipelines, multi-step tool use, and agent orchestration logic.
• Drive architectural decisions for the AI agent platform with a focus on fault tolerance, scalability, and extensibility.
• Build and optimize evaluation, observability, and classification pipelines that improve agent performance over time.
• Lead prompt engineering strategy for diverse customer use cases, including designing A/B testing and automated regression pipelines.
• Improve retrieval logic and embedding models to push RAG pipeline performance further.
• Monitor AI research and actively bring relevant techniques into production systems, using agentic CLIs and automation tooling to accelerate iteration cycles.
• Guide junior engineers and interns through code reviews, architectural discussions, and technical planning.
• Partner with PMs, designers, and sales engineers to translate real customer needs into well-scoped technical solutions.

• Prior experience with real-time or conversational interfaces.
• Familiarity with vector databases, search technologies, or large-scale data processing.
• Contributions to open-source LLM or AI tooling ecosystems.
• Native-level English fluency.

• Silicon Valley's equity program (1-year cliff) 
• Hybrid work policy, flexible work hours
• Be Your Best Self: 3.9 million won (prorated by start date) for expenses ranging from professional development classes and training, to personality assessments, to gym memberships, to books, to fitness classes, to mental health services, to massages
• Learn a Language benefit - up to 3.6 million won per year towards language lessons 
• Monthly team building cost
• Partial support for commuting costs
• Free parking at the Seolleung office
• Group insurance support that covers employees, spouses, and children
• Medical checkup support, including MRI and genetic testing - for the employee and one family member
• Seven additional paid holidays in addition to annual leave (Boost leave, Birthday leave, etc.)
• Unlimited snack bar filled with snacks, beverages, and instant noodles every day
• 12-week paid parental leave support (available for both mothers and fathers)
• Other support programs for congratulations and condolences

We offer a flexible work schedule at Sendbird.  We also value collaboration and relationship building.  With those values in mind, we require all employees to gather with their team in the office three days per week as a minimum.  Some of our roles require a more frequent in-office schedule.  Please work with your manager to understand the office time requirements for your position.

There is no such thing as a perfect candidate and the best employees come from a wide range of backgrounds, experiences, and skill sets. Sendbird is a place where everyone can learn and grow. We respect, promote, and encourage diversity for equal employment opportunities and encourage you to apply if this role excites you.

We're not adding AI to a legacy product. We rebuilt from the ground up with a decade of proprietary conversation data that most AI companies would kill for. The people here are moving with urgency because the opportunity is real and the window is now. If you want to own something that matters and see it ship to millions of users, this is the place.