AI Engineer III

Gopuff

72d • $215,000 - $275,000

About The Position

We’re reimagining instant shopping so technology never stands in the way, it accelerates you exponentially toward your goal by forming a deep connection with your needs and desires. In the Personal Superintelligence Lab, you will lead the design and deployment of agentic AI that reasons over rich, real‑world context and constraints, grounded in up‑to‑the‑minute knowledge and leveraging our unparalleled delivery speed. Your work will push the state of the art in alignment, grounding, and multi‑agent orchestration—while landing breakthroughs safely and at scale in production.

Requirements

• PhD in Computer Science, Machine Learning, or equivalent research experience with significant contributions to AI/ML literature.
• 7+ years of building and shipping large‑scale ML systems with significant ownership; proven impact in production LLM or RL‑driven products.
• Mastery of advanced fine-tuning techniques including LoRA/QLoRA, adapter methods, and parameter-efficient transfer learning.
• Research experience with agentic AI frameworks, multi-agent systems, and declarative programming approaches.
• Strong systems engineering capabilities with PyTorch, distributed training, and cloud-native ML infrastructure.
• Track record of publications in top-tier venues or equivalent industry impact.
• Deep expertise in transformer architectures, SFT, and RLHF; hands‑on leadership with RLVR and verifiable reward design.
• Mastery of policy optimization and the ability to extend/regularize policies under safety, latency, and cost constraints.
• Strong grounding in offline evaluation, counterfactual estimators, and safe online ramp strategies.
• Systems fluency: PyTorch, distributed training, low‑latency serving, observability, and cloud‑native ML infra.
• Demonstrated leadership across cross‑functional teams, with clear communication and mentoring track record.
• Commitment to responsible AI: privacy, safety, and alignment principles embedded end‑to‑end.

Nice To Haves

• Research or applied work in multi‑agent systems, decision theory, or declarative programming.
• Experience with formal methods for safety, program synthesis, or automated reasoning.
• Contributions to open‑source AI frameworks or foundational model development.
• Experience with privacy‑enhancing technologies, federated/on‑device learning, or identity/memory architectures.

Responsibilities

• Define the architecture, standards, and evaluation strategy that connect research to real‑world lift.
• Mentor colleagues and influence cross‑functional roadmaps.
• Ship systems that deliver measurable improvements to core customer and business outcomes.
• Set the strategy for context engineering to maximize precision/recall of key order metrics across sessions, households, locales, and time.
• Architect multi‑modal context integration and real‑time grounding with dynamic constraint satisfaction.
• Establish retrieval freshness, geo/time‑aware constraints, and memory policies; formalize context schemas and data contracts.
• Champion declarative prompt/program compilation for systematic, testable LLM behavior.
• Design multi‑agent orchestration patterns that yield robust emergent reasoning.
• Lead supervised reasoning-centered fine‑tuning with rigorous data curation, synthetic data generation, and QA.
• Own the reasoning architecture and evaluation strategy to deliver robust, low-latency, grounded outcomes at scale.
• Drive parameter‑efficient adaptation strategies with clear criteria for when to specialize vs. generalize.
• Architect RLHF and RLVR pipelines; build preference data loops, scalable oversight, and guardrails.
• Own policy optimization strategy with formal safety considerations.
• Ensure robust offline‑to‑online correlation via counterfactual/IPS/DR estimators and stress tests across traffic segments.
• Establish interpretability, controllability, and alignment verification practices for agentic systems.
• Develop safeguards against reward hacking and unsafe exploration; enforce distributional robustness and content policy compliance.
• Advance privacy‑preserving methods with privacy‑by‑design.
• Architect low‑latency, cost‑efficient inference with resilient fallbacks and red‑teaming.
• Build eval frameworks that tightly couple offline metrics with online performance and safety criteria; define promotion gates.
• Use relevant APIs to perform high‑fidelity data augmentation that strengthens grounding, disambiguation, and availability‑aware suggestions.
• Partner closely with Engineering and Data Science to design experiments, define success criteria, and iterate quickly from signal to lift.
• Translate ambiguous product goals into crisp technical milestones; maintain clear documentation, incident response, and learning playbooks.
• Mentor colleagues; raise the bar on design quality, reproducibility, and ethical rigor.

Benefits

• Medical/Dental/Vision Insurance
• 401(k) Retirement Savings Plan
• HSA or FSA eligibility
• Long and Short-Term Disability Insurance
• Mental Health Benefits
• Fitness Reimbursement Program
• 25% employee discount & FAM Membership
• Flexible PTO
• Group Life Insurance
• EAP through AllOne Health (formerly Carebridge)

What This Job Offers

• Job Type: Full-time
• Career Level: Senior
• Education Level: Ph.D. or professional degree