Staff Software Engineer

**Software Engineer — Bengaluru, India** **About Auxia** Auxia is an agentic AI platform that helps enterprises deliver personalized 1:1 customer journeys. Our Decision Agent determines the optimal message, timing, channel, and incentive for each individual user — processing 3B+ events/day , 25K+ queries/second , and 1B+ decisions/day at sub-100ms latency.

We serve global enterprises including Atlassian, The Guardian, Docomo, Comcast, Mercari, and MUFG Bank across email, push, in-app, and messaging channels. Backed by $23.5M from VMG Technology Partners, Stage 2 Capital, and MUFG Innovation Partners.

——————————————————————————————————————————————

The Hard Problems Auxia isn’t a typical SaaS platform. Here’s what makes the engineering genuinely interesting:

• Multi-tenant ML serving at scale. A single shared platform serves 20+ enterprise customers, each with millions of users, hundreds of treatments, and unique business constraints — all at sub-100ms p99. Every architectural decision has to balance isolation, performance, and cost across customers.
• Real-time decisioning under uncertainty. Our Decision Agent picks the best action for each user from hundreds of options using contextual bandits and ML models, learning continuously from live interactions. Cold-start, exploration-exploitation tradeoffs, and feedback loops are daily problems.
• Enterprise data integration. Every customer brings different data formats, volumes, schemas, and infrastructure (Snowflake, BigQuery, S3, CDPs). Onboarding a new customer’s data pipeline needs to be fast and reliable — we’re building toward fully automated ingestion.
• Agentic AI systems. Our Analyst Agent autonomously queries data, runs analyses, and generates insights for enterprise marketers. Building reliable, observable AI agents that interact with real production data is a frontier problem.

What You’ll Work On? Depending on your interests and strengths, you’ll work across some combination of:

• Platform Infrastructure — Kubernetes orchestration on GCP, service mesh, deployment automation, observability (metrics, tracing, alerting), cost optimization across a multi-region platform.
• Data Systems — High-throughput ingestion pipelines (Apache Beam/Dataflow, Pub/Sub), BigTable and BigQuery at terabyte scale, real-time feature stores, data warehouse and reporting infrastructure.
• Backend Services — Kotlin/gRPC microservices, treatment recommendation and scoring engines, experiment framework, configuration management, multi-tenant authorization.
• ML Infrastructure — Model training pipelines (Metaflow), model serving, feature engineering automation, A/B test evaluation, bandit algorithms in production.
• Frontend & Developer Experience — Next.js admin console, internal tooling, developer productivity, CI/CD pipeline optimization.
• Customer Integration — Forward-deployed engineering to onboard enterprise customers, building SDKs and integration tooling.

Our Tech Stack

• Languages: Kotlin (primary backend), TypeScript/Next.js (frontend), Python (ML pipelines)
• Infrastructure: GCP, Kubernetes, Docker, Terraform
• Data: BigTable, BigQuery, Apache Beam/Dataflow, Pub/Sub, PostgreSQL
• Services: gRPC/Protobuf, Spring Boot
• ML: Metaflow, custom bandit/scoring frameworks
• Tooling: Gradle, GitHub Actions, Linear, Figma

How We Work

• AI-native development. We use Claude Code extensively — for code generation, architecture exploration, code review, debugging, and documentation. Engineers here ship faster because they’re fluent with AI-assisted development. We’re building internal AI agents to automate parts of the DS and engineering workflow. If you’re excited about working at the intersection of building AI products and using AI to build, this is the place.
• Small team, high ownership. ~30 engineers across US, India, and Japan. No layers between you and production. You’ll own systems end-to-end — design, build, deploy, monitor.
• Onsite. We believe the best engineering happens in-person, especially at our stage. Fast iteration, whiteboard sessions, and same-room debugging.
• Ship weekly, not quarterly. Ideas go to production in days, not months. We make smart tradeoffs between speed and quality — and we trust engineers to make those calls.

What you bring?

• Strong fundamentals in distributed systems, data structures, and system design with 9-12 years of experience.
• Experience building and operating pro…