Staff Engineer, Core ML

Circuit•Austin, TX

About The Position

As a Staff Engineer, Core ML at Circuit, you’ll join Product Engineering and own the machine-learning capabilities at the center of the product: understanding complex technical documents and keeping long-running agents that work with them reliable. Circuit’s customers use blueprints, CAD files, specification sheets, and scanned PDFs containing text, vector geometry, and raster content. You will build models and algorithms that turn these documents into structured information, and tools that let agents use it. You will also improve how agents plan, choose tools, recover from failures, and behave consistently across runs. The role focuses primarily on algorithms and models rather than individual customer projects. Product engineers and forward-deployed engineers bring customer-facing failure patterns; you analyze model and agent performance, identify where algorithms, data, or training can improve results, and set evaluation standards to assess changes. You may join customer conversations when a problem requires your expertise. This is an applied engineering role focused on adapting and combining proven approaches, measuring them against real customer data, and shipping them to production. Success is measured by production impact rather than published papers. The existing product includes a Go-based long-running agent loop driven by frontier model APIs, specialized retrieval, extraction, and execution tools, and open-source and internally fine-tuned vision-language models hosted on KServe.

Requirements

Hands-on AI/ML engineering experience and a track record of shipping models in production.
Experience training and fine-tuning production models, including building the datasets and pipelines behind them.
Experience building and operating ML evaluation pipelines, including datasets, metrics, and regression testing for production models and agents.
Experience profiling data distributions, identifying gaps and bias, and curating high-quality training and evaluation datasets.
Experience with vision and multimodal models, particularly for technical documents such as PDFs and CAD containing text, vector geometry, and raster content.
Strong knowledge of data structures, algorithms, and software engineering principles.
Proven expertise with LLMs, deep learning systems, and applied ML.
Experience debugging production ML systems with AI observability tools, such as Langfuse or Opik, and tracing failures to their root causes.
Proficiency in Python and relevant ML frameworks, such as PyTorch and vLLM.
Daily, fluent use of AI coding tools, including writing effective prompts, iterating on incorrect output, and distinguishing working code from code that only appears correct.
Ownership mindset focused on delivering working production systems.
Strong product judgment and understanding of how system behavior affects user experience and trust.
Ability to review agent instructions, skills, and documentation for clarity, precision, and reliability.
Curiosity about system behavior in practice and how customers use the product.
Ability to drive ambiguous problems to working solutions and bias toward shipping.
Judgment about when work is ready to ship and willingness to prioritize customer outcomes.
Clear communication of ML tradeoffs to engineers, FDEs, and product stakeholders, with progress shared early.
Ability to critically evaluate AI output and catch hallucinated APIs, incorrect logic, and subtle bugs.
Systems thinking, including attention to correctness, determinism, and failure modes in long-running agents.
Rigor in measurement, including distinguishing real improvements from evaluation noise.
Ability to raise the team’s standards through design reviews, code reviews, and mentoring.
Low-ego, team-oriented approach.

Nice To Haves

Deep expertise in at least one of the following: agent harness design and execution; document extraction and structured data understanding; high-resolution computer vision; VLM fine-tuning and inference; retrieval systems, including embeddings and reranking; or agent evaluation and optimization.
Computational geometry experience or experience with vector drawing formats such as CAD and vector PDFs.
Go experience.
Familiarity with construction, civil engineering, or related industries where complex technical drawings and specifications are central.

Responsibilities

Develop algorithms to extract meaning from complex technical documents containing text, vector geometry, CAD, and raster content, using vision models, geometric algorithms, and language understanding.
Integrate document-understanding capabilities into tools used by Circuit’s long-running agentic systems.
Improve the reliability, determinism, and effectiveness of agentic systems, including planning, tool use, and failure recovery.
Build and extend training and evaluation pipelines for core models and agents, measuring changes against real customer data before release.
Analyze production performance, identify where and why algorithms fail, and develop improvements.
Identify dataset gaps and bias, and address them through curation, labeling, or synthetic data.
Track new research and techniques and bring promising approaches into production.
Mentor and elevate other engineers through technical leadership, reviews, and guidance.