Senior AI Developer

Cadstrom•Montreal, QC

About The Position

We are hiring a Senior AI Developer who is passionate about electronics. You'll play an early and foundational role building the AI tools at the heart of Cadstrom – from LLM-based pipelines that extract structured component data from manufacturer datasheets, to AI-driven validation of customer circuits, to self-improving systems that get smarter every time they run. The work is detail-dense and multi-modal: graphs, tables, cross-referenced text, and parameters buried deep in application notes. You'll drive our experimentation framework end-to-end, dig deep into failure modes, and ship incremental improvements that compound over time. We move fast – the product is shipping, customers are live, and every week matters. This is a high-ownership role for a detail-obsessed AI developer who treats Cadstrom's outcome as their own. What are we doing? We are accelerating the world's hardware ingenuity by blending generative AI, ML, and novel deterministic approaches to electronics simulation and validation. We're providing our enthusiastic customers intelligent and automated access to the most advanced validation tools for catching design errors before they reach fab – helping engineers of all levels deliver their designs faster, cheaper, and with less waste. Who are we? We are a small team of innovators who are passionate about fundamentally improving how electrical engineers deliver electronics with higher confidence. We are backed by Bison Ventures, Innovation Endeavors, and the Allen Institute for AI's Incubator. Our distinguished founders are responsible for creating the largest product development firm in the US, and building Azure Quantum (and more!) at Microsoft. We know you'll be in good company.

Requirements

3-4 years of hands-on experience in AI/ML or applied data engineering with shipped work in production or active pilot.
Practical experience with LLMs (prompt engineering, few-shot/ICL, RAG, or agentic patterns) and interest in multi-modal inputs (PDFs, tables, figures).
TypeScript experience – typing, packaging, services, and automated tests.
Strong grasp of experiment design: reproducibility, significance, appropriate metrics, and ground-truth curation.
Git proficiency and collaborative dev habits (peer reviews, readable self-documenting code).
Meticulous attention to detail in debugging and root-cause analysis.
Excellent written and visual communication; you summarize messy problems crisply and use plots and figures to convey insights.
Undergraduate degree in Computer Science, Engineering, or a related STEM field.

Nice To Haves

Experience building and optimizing tools (MCP-based or otherwise) for use by AI agents.
Electronics fundamentals (e.g. interpreting a datasheet) via coursework or hobbyist experience.
Retrieval (vector DBs), chunking/embedding strategies, and context budgeting.
LLM-ops platforms (LangSmith, LangFuse, DSPy).
Data pipelines in a cloud environment (AWS/GCP/Azure).
Startup experience, or a research-based Master's in a STEM field.
Based in Montreal or the GTA

Responsibilities

Manage highly automated AI flows for data extraction from electrical component datasheets.
Build AI pipelines that perform validations on customer circuits, integrating retrieval, agentic systems that think like a seasoned EE, and deterministic checks into trustworthy end-to-end flows.
Build benchmarking pipelines and evaluation tooling that electrical engineering domain experts use to analyze, calibrate, and improve our AI systems.
Design and ship self-improving systems exposed to our customers – closing the loop between production signal, evaluation, and model and prompt improvements.
Build and optimize tools (MCP-based or otherwise) for AI agents, ensuring agents can reliably interact with Cadstrom's platform and data.
Diagnose failure modes by digging into model outputs and traces, documenting root causes, and proposing fixes at the data, prompt, or schema level.
Partner closely with electrical engineers to internalize domain conventions and translate that understanding into better prompts, schemas, and evaluations.