Senior AI Engineer

About thumpN
thumpN is an AI-native live events discovery and ticketing platform. We power everything from club nights to large festivals and event pages, ticketing to gate entry, with AI at the core of how fans discover events. Shadow is our agentic assistant that helps fans find events and book tickets in conversation. This role builds it.
Why This Role Exists
You’ll own the engineering behind Shadow. You’ll make sure it understands what a fan wants, finds the right events, holds a good conversation, and keeps fans coming back. You’ll shape the architecture, write a significant amount of the code yourself, and build the evals and monitoring needed to know whether it’s actually working. This is an ownership role for someone who can take Shadow from technical direction to production with minimal oversight.
Key Responsibilities
Design and build the agent end to end: orchestration, tool calling, multi-turn state and memory, retrieval, and guardrails
Build the core loop: understand intent, find the right events, and help fans book when they're ready
Handle the messy parts of conversation: multilingual queries, session handoffs, and recovering well when the agent gets it wrong
Define what "good" looks like for each capability, build eval sets from real sessions, and use them to decide what ships
Set up tracing and debugging so bad responses can be traced back to a prompt, a tool, retrieval, or code
Choose models and architecture based on latency, quality, and cost — including knowing when a smaller model or a simple rule does the job
Raise the quality of engineering around you through reviews and example
Work with product and leadership on what Shadow does next
What we are looking for
7+ years of software engineering experience, with at least 1+ year on agentic systems
You've shipped an agent that takes actions, not just one that answers questions
Hands-on with the current stack: model APIs, orchestration, RAG, structured outputs, and multi-turn state
You can explain how you measured whether something you shipped actually worked
Strong backend skills in Python and/or TypeScript
You think about latency, cost, prompt versioning, and what happens when things fail
You've taken ambiguous projects from start to finish
You use AI tools well in your own work and know where they fall short
Nice to have
LLM observability tools (Langfuse, LangSmith, Datadog LLM Observability)
Payments, ticketing, or marketplace experience in India
Fine-tuning or serving open-weight models
Familiarity with our stack: Next.js, Go, AWS, Cloudflare

Work arrangement
No