Machine Learning Engineer

Migaru AI builds an AI-native CRM that reads a sales team's email, calendar and calls, files each conversation against the right account and deal, and keeps the record current without manual data entry. We're looking for a Machine Learning Engineer to help build and improve the models and pipelines that power this system, from extracting structured facts out of unstructured conversations to generating accurate, useful drafts for review.

What you'll do

• Design, train, and evaluate models for tasks such as information extraction, summarization, classification, and ranking over email, calendar, and call transcript data
• Build and maintain data pipelines for training and evaluation, including labeling workflows and quality checks
• Take models from prototype to production, working closely with backend engineers to ship reliable, low-latency inference
• Monitor model performance in production, diagnose failure modes, and iterate based on real usage
• Collaborate with product on where human review and approval belong in the workflow, given model confidence and error costs
• Evaluate and integrate third-party LLMs and APIs where they're a better fit than custom models

What we're looking for

• Professional experience building and shipping machine learning systems in production
• Strong Python skills and familiarity with standard ML/NLP tooling (e.g. PyTorch, Hugging Face transformers)
• Experience working with large language models, including prompting, fine-tuning, or retrieval-augmented approaches
• Solid understanding of evaluation methodology: how to measure quality, catch regressions, and avoid fooling yourself with bad metrics
• Comfort working with messy, real-world text data (emails, transcripts, meeting notes)
• Clear communication, especially when explaining model behavior and limitations to non-ML teammates

Nice to have

• Experience with named entity recognition, relation extraction, or document understanding
• Experience building systems that must handle ambiguity gracefully, such as deduplication or entity resolution
• Familiarity with speech-to-text or working with call/meeting transcript data
• Experience designing human-in-the-loop review workflows for ML-generated output