AI Conversational Session Summarization Developer

Mevra Labs

Remote · $100k – $130k / year · Full-time · Posted 4 months ago

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

About the role

A long conversation history is both an asset and a liability. The asset: rich context about what the user has said, tried, and decided. The liability: every token costs compute, and a 50-turn conversation might contain 45 turns of context that is now irrelevant and 5 turns of information that is critical to the current task. We build conversational AI infrastructure for enterprise clients. Our session summarization layer compresses conversation history into structured summaries that preserve the decision-relevant content while dramatically reducing token count for subsequent turns. The challenge is selectivity: summaries that are too aggressive lose critical nuance; summaries that are too conservative do not solve the cost problem. You will build the summarization system: designing the relevance scoring that identifies which turns to preserve verbatim versus summarize versus discard, implementing the structured summary format that downstream models can reliably extract from, and building the evaluation framework that measures summarization quality against human-labeled ground truth.

Fully remote.

Responsibilities

Design relevance scoring for conversational turn-level preservation decisions
Build structured summary format optimized for downstream model extraction
Implement evaluation framework measuring summarization quality against human labels
Develop context-sensitive compression that adapts to session topic and user intent
Build A/B testing infrastructure comparing summarized vs full context performance
Optimize summarization latency for real-time conversation serving

Requirements

3+ years in NLP engineering with summarization, dialogue, or LLM systems experience
Experience with abstractive summarization and structured text generation
Familiarity with LangChain, Hugging Face, or similar NLP tooling
Strong Python and API development skills
Understanding of token economics and context window management in LLM applications