Founding ML & Data Architect

ObjectiveIf you've ever stopped to ask what it actually looks like when AI starts acting in someone's physical life, not just actuators and answers but a breadth of embodied, agentic intelligence cohabiting with people across the places where they spend their time, and you didn't flinch at either the implications or the opportunity, you're the kind of person we want to talk to.Tethral, Inc. is building an enabling layer that allows enterprises to safely and securely deploy physical AI experiences to consumers at scale. The company is starting with an application that lets consumers connect software and hardware and demonstrate how they want physical AI to fit into their daily lives.Core ResponsibilitiesData architecture: end-to-end schema for multi-modal behavioral capture, structured so raw interaction becomes state-action sequences usable as training data for generative models. Data shape is a foundational constraint, not downstream cleanup.Substrate engineering: systems that transform raw sensor and interaction data into structured training data under strict consent and provenance frameworks.Validation systems: measurement-based evaluation that the representation is generative and generalizes, supporting held-out prediction and counterfactual rollouts against out-of-distribution reality, not training metrics.Data asset ownership: the structural decisions that shape the behavioral substrate, weighted to its long-term value over short-term application features.Technical RequirementsDesign high-throughput data schemas and pipelines where the shape of the data is a foundational constraint, and defend those structural choices against alternatives.Convert high-dimensional, noisy sensor data (wearables, prosthetics, or robotics) and messy human interaction data into structured, high-fidelity datasets.Representation learning in practice: self-supervised methods and learned embeddings, with the judgment to choose an approach and defend it.Sequence and temporal modeling, specifically state-space and latent-variable formulations. You model behavior as states evolving over time.You have built or trained generative or model-based systems from the ground up, not only consumed or fine-tuned someone else's. This is a hard requirement and is verified in the process.Measurement-based validation: a track record of proving a model generalizes rather than fits.Advanced degree or equivalent chops shipping production grade systems. Compensation & StructureEquity-first. Work begins immediately.No cash compensation until the round closes. The raise is expected to close in approximately two months, at which point cash begins.At full velocity pay is $200,000 USD, with equity and benefits. Founding-member equity, 2 to 4 percent, standard vesting. Band position reflects commitment and risk carried through the raise.Co-founder-level stake is possible commensurate with experience and riskEquity begins following a short work trial with vesting schedule ApplicationShipping record: a production-grade system or feature you deployed. State impact and the technical trade-offs.AI-augmented workflow: how you use AI in engineering, across three points: how, for what, and where you can and cannot assess output quality.Technical proof: a repository link, a technical teardown, or publication.Signal: the last hackathon or build event you took part in, and what you made.Detail the generative model or system you built. We strongly encourage applications from people of color, candidates across all age groups, people with disabilities, neurodivergent candidates, LGBTQIA+ candidates, and veterans. Please note we cannot sponsor Work Visa’s at this time. Reach out directly , info@tethral.ai.