{"schemaVersion":"jobsearcher.job.v1","id":"f6eed76dd0864e596a88fbc2","url":"https://jobsearcher.com/jobs/f6eed76dd0864e596a88fbc2","canonicalUrl":"https://jobsearcher.com/jobs/f6eed76dd0864e596a88fbc2","title":"ML Research Engineer","description":"ML Research Engineer.Train differentiable models that discover exact discrete programs and circuits from examples, not approximations.\r\nLearn exact programs, not approximate functions.We train models that learn the exact discrete program behind a set of examples, until the model itself converges into an exact digital circuit. This is not predictive ML. There is no acceptable error bar. A model that's 99% right is wrong, because the goal is a circuit that computes the function, not one that approximates it. (Mapping that converged circuit onto real FPGA fabric is the compiler team's job; you get it to converge.)\r\nYou'll work where deep learning, program synthesis, and logic meet: differentiable relaxations of discrete program search, optimization methods that drive a soft model toward a hard one (hardening, sparsification, generalization), and recurrent or stateful architectures that learn exact update rules and collapse cleanly into discrete logic.\r\nThe research question is sharp and wide open: can a network be trained until it doesn't approximate a circuit, but becomes one?\r\nWhat you'll do.Design sharp experiments and run rigorous ablations.\r\nDiagnose whether failures come from architecture, objective, optimization, or data generation.\r\nBuild minimal tasks that isolate missing primitives or missing inductive bias.\r\nImplement new training methods for discrete and near-discrete models.\r\nAnalyze soft-vs-hard mismatches and propose ways to close them.\r\nShape architectures that are both trainable and that collapse into exact, discrete logic.\r\nMaintain a high-quality experimental codebase in PyTorch.\r\nWhat you'll bring.Strong PyTorch and practical deep learning engineering.\r\nStrong grasp of optimization, gradient flow, and training instability.\r\nStrong experimental discipline: clean baselines, controlled ablations, reproducibility, and reading negative results honestly.\r\nA bias toward small, decisive experiments over big, inconclusive ones.\r\nStrong coding in Python, and comfort reading lower-level code when needed.\r\nNeural program synthesis or algorithmic reasoning.\r\nFormalizing problems as search over discrete structures.\r\nA feel for when reaching zero training loss really means the model has collapsed into an exact circuit, not a soft approximation that only looks discrete.\r\nDesigning synthetic tasks and curricula.\r\nSequence models, recurrent state, or memory-based models.\r\nSAT/SMT, combinatorial optimization, or search.\r\nLogic synthesis, circuits, compilers, or PL.\r\nDifferentiable relaxation methods for structured prediction.\r\nA screenshot of your Factorio megabase\r\nBoolSi is an equal opportunity employer. All qualified applicants will receive consideration without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, disability, or veteran status.#J-18808-Ljbffr","company":"Boolsi","rawCompany":"boolsi","city":"Brooklyn","state":"NY","isRemote":false,"isActive":false,"createdAt":"2026-10-10T01:32:50.407Z","occupations":[{"code":"15-1221.00","title":"Computer and Information Research Scientists","slug":"computer-and-information-research-scientists"},{"code":"15-1252.00","title":"Software Developers","slug":"software-developers"},{"code":"15-1299.08","title":"Computer Systems Engineers/Architects","slug":"computer-systems-engineers-architects"}],"industries":[{"code":"541715","title":"Research and Development in the Physical, Engineering, and Life Sciences (except Nanotechnology and Biotechnology)","slug":"research-and-development-in-the-physical-engineering-and-life-sciences-except-nanotechnology-and-biotechnology"},{"code":"541511","title":"Custom Computer Programming Services","slug":"custom-computer-programming-services"},{"code":"541512","title":"Computer Systems Design Services","slug":"computer-systems-design-services"}],"jobPosting":{"@context":"https://schema.org","@type":"JobPosting","title":"ML Research Engineer","description":"ML Research Engineer.Train differentiable models that discover exact discrete programs and circuits from examples, not approximations.\r\nLearn exact programs, not approximate functions.We train models that learn the exact discrete program behind a set of examples, until the model itself converges into an exact digital circuit. This is not predictive ML. There is no acceptable error bar. A model that's 99% right is wrong, because the goal is a circuit that computes the function, not one that approximates it. (Mapping that converged circuit onto real FPGA fabric is the compiler team's job; you get it to converge.)\r\nYou'll work where deep learning, program synthesis, and logic meet: differentiable relaxations of discrete program search, optimization methods that drive a soft model toward a hard one (hardening, sparsification, generalization), and recurrent or stateful architectures that learn exact update rules and collapse cleanly into discrete logic.\r\nThe research question is sharp and wide open: can a network be trained until it doesn't approximate a circuit, but becomes one?\r\nWhat you'll do.Design sharp experiments and run rigorous ablations.\r\nDiagnose whether failures come from architecture, objective, optimization, or data generation.\r\nBuild minimal tasks that isolate missing primitives or missing inductive bias.\r\nImplement new training methods for discrete and near-discrete models.\r\nAnalyze soft-vs-hard mismatches and propose ways to close them.\r\nShape architectures that are both trainable and that collapse into exact, discrete logic.\r\nMaintain a high-quality experimental codebase in PyTorch.\r\nWhat you'll bring.Strong PyTorch and practical deep learning engineering.\r\nStrong grasp of optimization, gradient flow, and training instability.\r\nStrong experimental discipline: clean baselines, controlled ablations, reproducibility, and reading negative results honestly.\r\nA bias toward small, decisive experiments over big, inconclusive ones.\r\nStrong coding in Python, and comfort reading lower-level code when needed.\r\nNeural program synthesis or algorithmic reasoning.\r\nFormalizing problems as search over discrete structures.\r\nA feel for when reaching zero training loss really means the model has collapsed into an exact circuit, not a soft approximation that only looks discrete.\r\nDesigning synthetic tasks and curricula.\r\nSequence models, recurrent state, or memory-based models.\r\nSAT/SMT, combinatorial optimization, or search.\r\nLogic synthesis, circuits, compilers, or PL.\r\nDifferentiable relaxation methods for structured prediction.\r\nA screenshot of your Factorio megabase\r\nBoolSi is an equal opportunity employer. All qualified applicants will receive consideration without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, disability, or veteran status.#J-18808-Ljbffr","datePosted":"2026-10-10T01:32:50.407Z","dateModified":"2026-10-10T01:32:50.407Z","hiringOrganization":{"@type":"Organization","name":"Boolsi","sameAs":"https://jobsearcher.com"},"jobLocation":{"@type":"Place","address":{"@type":"PostalAddress","addressLocality":"Brooklyn","addressRegion":"NY","addressCountry":"US"}},"identifier":{"@type":"PropertyValue","name":"JobSearcher","value":"f6eed76dd0864e596a88fbc2"},"url":"https://jobsearcher.com/jobs/f6eed76dd0864e596a88fbc2"}}