Software Engineer III

ICONMA is seeking a Software Engineer III for their remote team. The role involves designing datasets for training machine learning models, running evaluations, and developing data-heavy internal tools.

Responsibilities:

• Design human and model-based datasets for training and evaluating MLLMs (image/video/audio/text data)
• Run frontier model evaluations and implement common metrics
• Use MLLMs to bootstrap, augment, and rebalance datasets
• Fine-tune models on labeled datasets
• Ship annotation task UIs and data-heavy internal tools (React/TypeScript)
• Run secure ingestion pipelines from warehouse/warm-storage/flat files at scale
• Meeting deadlines, executing tasks with accuracy, etc

Requirements:

• ML/AI fundamentals - fine-tuning (SFT/preference), prompting, model-based eval, and the failure modes of each
• At least one year of hands-on work in the field. Multimodal dataset experience - building image/video/audio/text datasets to train or evaluate models
• Strong Python + ML tooling - PyTorch (or equivalent), HF ecosystem, comfort running training/sampling jobs on real infra
• Web/UI engineering - production React/TypeScript; design, vibe code and ship clean, usable interfaces for annotation and internal tools
• Data pipelines - SQL plus reliable large-scale data movement: batching, idempotent dedupe, parallelism, retries
• Experience in mid sized tech, preference for (Google, Apple, client, or other FAANG)
• 5-10+ years of hands-on experience in building multimodal datasets, implementing machine learning evaluations, and developing production-level data and annotation tools
• Academic background in Computer Science, Engine
• MS or PhD in AI or related field
• Research track record in MLLMs - demonstrated through publications or impactful contributions to major projects
• Open-source contributions - experience shipping and maintaining public-facing repositories