Principal Signal Processing /… at Foresite Labs… · San Diego

Location: San Diego, CA | Full-Time | Salary: $260,000 – $270,000 Position Overview We are building a high-throughput analysis system for noisy, information-rich measurement data. The work sits at the boundary of signal processing, statistical estimation, physical-system modeling, and production software. The core challenge is turning an ambiguous real-world signal into a calibrated algorithm that can be trusted under operational constraints. This role owns algorithmic formulation and delivery. You will identify the right model for the signal, define the measurements that matter, build the algorithmic path from prototype to production, and work with partner teams when the data shows that the physical system or measurement process needs to change. The strongest fit is a hands-on principal engineer who can sit with an imperfect signal, derive the right question, defend the math, write production-intent code, and explain the tradeoff in plain language to people outside their specialty. Roles and Responsibilities - Formulate and ship algorithms for noisy measurement data, including estimation, detection, calibration, confidence modeling, drift analysis, and systematic-error reduction - Build modeling and analysis frameworks that explain current performance, identify the factor limiting data quality, and prioritize the next experiment or engineering change - Use simulation and controlled datasets to make algorithm work falsifiable: isolate failure modes, bound achievable performance, and separate model error from measurement error - Work directly with hardware, measurement, and production-engineering partners on what to measure, what to change, and how to tell whether a change improved the system - Write production-intent code with clear interfaces, deterministic behavior, and tests grounded in measured or simulated truth - Translate algorithmic insight into product and system-design decisions without turning every discussion into a research project What We're Looking For (Must-Haves) - First-principles problem formulation. Given a noisy, partly characterized signal and an incomplete goal, you can derive the right question before reaching for tools. You can defend an SNR estimate, likelihood model, error budget, or bias/variance tradeoff from the ground up. - Shipped algorithmic ownership. You have personally taken an algorithm in signal processing, communications, radar, computational sensing, imaging, controls, physical-layer systems, or a comparable domain from problem statement to a deployed system other people depend on. - Production toolchain fluency. You are strong in Python for analysis and prototyping, and you can write production-intent Rust or modern C++ where correctness, speed, memory layout, and maintainability matter. MATLAB-only experience is not enough for this role. - Cross-disciplinary collaboration. You have shaped measurement, hardware, or physical-system decisions using algorithmic evidence, not just consumed data handed to you by another team. - Curiosity about the source of the data. You want to understand what the system actually measured, how the data was produced, and which assumptions are safe enough to encode. Education - MS or PhD in EE, CS, Physics, Applied Math, Statistics, or related field or equivalent demonstrated work - MS + 14 years, or PhD + 10 years, in DSP, image analysis, communications, physical-layer algorithms, computational imaging, or comparable. What matters is demonstrated depth and shipped ownership Strongly Preferred - Rust at shipping depth for performance-critical algorithm or numerical components - Communication-theory or estimation-theory tools applied outside their original domain: maximum-likelihood detection, decision feedback, adaptive filtering, channel modeling, state estimation, or related methods - GPU-accelerated algorithm work in a real performance regime, especially CUDA or comparable accelerator programming - Experience with simulation-driven algorithm development where the simulator and algorithm improve together - ML used as one tool inside a broader estimation or signal-processing framework, with clear understanding of where it helps and where it hides failure modes - Public or shareable evidence of depth: papers, patents, open source, technical talks, postmortems, or a concrete shipped system you can discuss Nice to Have - Physical measurement systems with non-trivial analysis pipelines - Calibration and confidence / quality modeling on production outputs - Custom FFI or systems-language boundaries for performance-critical numerical code - Experience helping non-algorithm teams use diagnostics without oversimplifying the underlying model We are an equal opportunity employer. We thrive on diversity and collaboration.