ML Engineer: Memory & Retrieval (Remote)

SendGrid, Inc.; United States; Remote; USD 128,000 - 170,000; Full time

Benefits offered by this job

Health insurance

401(k) retirement plan

Paid time off

Job summary

Twilio is seeking a Machine Learning Engineer for the Conversation Memory team. You will build scalable pipelines that convert raw conversation data into structured, searchable memory and maintain evaluation systems to ensure memory accuracy as products evolve.

You will work on the memory layer behind Twilio's Conversational AI platform, extracting meaningful data from millions of conversations, mapping it to customers, and retrieving it in milliseconds for agents.

Qualifications

• Relevant experience building backend services and picking up new languages/frameworks/tools.
• Experience designing or working with evaluation/quality-measurement systems (regression/quality metrics).
• Strong grounding in observability: logs, metrics, tracing used to debug distributed systems.

Responsibilities

• Build the extraction pipeline to turn raw data into structured memory with tuned prompts/schemas.
• Own the retrieval path using embeddings, vector search, and ranking to answer what we know about a person."
• Contribute to evaluation systems to measure memory accuracy and set benchmarks and gates.
• Diagnose production issues across ingestion, extraction, and indexing paths using logs and metrics.
• Collaborate with Product and ML teams to translate customer problems into system changes and evaluation criteria.

Skills

Backend development

Observability

Evaluation systems

AI tooling

Tools

Go

Python

Kafka

LLM tooling

Job description

Twilio is seeking a Machine Learning Engineer for the Conversation Memory team. You will build scalable pipelines that convert raw conversation data into structured, searchable memory and maintain evaluation systems to ensure memory accuracy as products evolve.

You will work on the memory layer behind Twilio's Conversational AI platform, extracting meaningful data from millions of conversations, mapping it to customers, and retrieving it in milliseconds for agents.