Founding ML Engineer — Rust, Models & Agent Systems

Loci; New York (NY); On-site; USD 180,000 - 220,000; Full time

Job summary

Loci is seeking a founding engineer to own core ML systems, from data handling and model training to local inference, tool use, memory, and retrieval. You’ll write Rust, reason about model boundaries, and shape product features with pragmatic engineering for reliable on-device AI.

You’ll move between model work and product engineering, decide when a smaller model suffices, and build end-to-end systems with high density of experimentation and responsible resource use.

Qualifications

• Experience training and fine-tuning language models and evaluating their performance.
• Strong Rust engineering skills, ownership and performance profiling.
• Experience with llama.cpp or local inference and quantized models.
• Experience building agent systems and tool-calling mechanisms.
• Excellent written and verbal communication; ability to explain complex tradeoffs.

Responsibilities

• Train, fine-tune, and evaluate models against product requirements with reproducible experiments.
• Design boundaries between models and the harness; decide responsibilities that require model judgment vs code.
• Build Rust software for inference, agent workflows, memory, and retrieval with strong type safety.
• Integrate models with llama.cpp and related engines; diagnose compatibility issues and optimize performance.
• Write and maintain system prompts, templates, and tool definitions; handle malformed calls and failures.
• Develop reproducible evaluation workflows to measure latency, reliability, and resource use.
• Document decisions and tradeoffs; communicate progress and blockers clearly.
• Mentor teammates on model and harness behavior; share research and best practices.

Job description

Loci is seeking a founding engineer to own core ML systems, from data handling and model training to local inference, tool use, memory, and retrieval. You’ll write Rust, reason about model boundaries, and shape product features with pragmatic engineering for reliable on-device AI.

You’ll move between model work and product engineering, decide when a smaller model suffices, and build end-to-end systems with high density of experimentation and responsible resource use.