PProfound
Profound

1-10 · Pre-Seed

LLM Engineer

Bengaluru|

Experience

0-3 years

Work Mode

On-site

About Profound

Most professional networks surface credentials. Profound is built on a different belief: that what a person can do matters more than where they have been, and that the right technology can make that visible at scale.

The product is an AI-native professional network designed to make human potential discoverable, built for a global market from day one.

Profound is at pre-seed, with a founding team that has scaled B2C products in India across Delhivery, Ola, Snapdeal, and Zomato.

About the Role

Build the LLM execution layer behind Profound.me’s AI representative. Improve output quality, reliability, response time and cost. Full-time, on-site in Bengaluru.

What You'll Own

  • Build model routing and fallbacks across providers using tools such as LiteLLM. Run and benchmark local or open-weight models with Ollama, vLLM or similar inference tools. Version prompts and build repeatable evaluations with tools such as promptfoo or DeepEval. Trace LLM requests and monitor quality, latency and cost using Langfuse, Phoenix or similar tools. Compare deployment options across clouds and turn production failures into regression tests.

Signs You'll Be A Great Fit

You have 1–3 years of engineering experience and hands-on experience building LLM applications. You can write production software, evaluate model outputs, debug failures, and explain quality, cost and latency trade-offs. Practical experience running open-weight models and cloud inference is valuable.

How We Work

Test prompts and models on real Profound.me agent tasks, not selected demo outputs. Check changes against versioned evaluations before release. Turn production failures into regression tests. Compare quality, latency and cost before changing models or routing rules. Follow requests end to end: context, prompts, routing, inference and output checks. Choose hosted APIs or self-hosted models based on measured results, not model popularity.

The Opportunity

Build the decision layer behind Profound.me’s AI representative: which model handles each task, with what context, and at what cost. This is not a prompt-writing-only role. Own evaluations, routing, model deployment and observability together. See how your choices change agent behavior, response time and operating cost.