Senior Staff AI Scientist

Uniphore•Palo Alto, CA

8d•$232,900 - $320,250•Onsite

About The Position

Uniphore is building the world's best-in-class Business AI platform to enable business users to leverage the advances of artificial intelligence to solve problems in their specific business processes. Uniphore provides a composable, sovereign, and secure solution to business users, and drives AI adoption in organizations with innovative solutions to building, serving, and optimizing these AI solutions. You will be working at the core of a category-defining product — building the agent learning platform and SLM AI flywheel that powers our Business AI cloud. This is the system that closes the loop: AI agents operate in production, capture real-world signal, and continuously improve through automated fine-tuning and evaluation. If you are ready to own the full ML stack behind agentic AI at enterprise scale — from orchestration architecture through model optimization, evaluation infrastructure, and production delivery — come join us.

Requirements

MS or PhD in Computer Science, Machine Learning, Statistics, or related field
8+ years designing, building, and operating production ML systems, with hands-on experience with frontier and open-source models
Proven track record owning agentic AI systems or closed-loop model improvement pipelines in production — not prototype quality
Deep experience with LLM or SLM fine-tuning: SFT, RLHF/DPO, data curation, and rigorous evaluation design
Experience translating business impact into quantitative metrics and designing statistically sound experiments
Track record of influencing technical decisions beyond your immediate team — through design docs, architectural reviews, or cross-functional alignment
Strong communication skills, verbal and written, with ability to present technical strategy to non-technical stakeholders

Nice To Haves

Experience designing evaluation frameworks for agentic systems (trajectory eval, task success, robustness benchmarks
Demonstrated influence at org or platform level: architectural standards or platform decisions that multiple teams built against
Familiarity with agentic orchestration frameworks ( LangGraph, or equivalent)
Background in enterprise NLP, conversational AI, or contact center / CX domains
Publications at top-tier peer-reviewed venues or significant open-source contributions in relevant areas
Experience at fast-growing companies or in agile, high-ownership engineering environments

Responsibilities

Design and implement production-grade agentic systems capable of multi-step reasoning, planning, tool use, and decision-making under real operational constraints (latency, cost, safety).
Own the orchestration layer of the agent learning platform: agent memory, inter-agent communication, failure recovery, and reliability patterns at enterprise scale.
Translate abstract product requirements into reliable AI behaviors and set the architectural standards the team builds against.
Own the closed-loop learning pipeline: capturing production signal from deployed agents, triggering fine-tuning cycles, and gating model promotion into production.
Fine-tune and adapt small and medium-sized foundation models using techniques such as PEFT, SFT, distillation, and reinforcement learning (RLHF, DPO).
Drive model selection decisions (SLMs vs. larger models) based on use-case requirements, latency SLAs, and empirical evidence.
Define and build evaluation strategy for agentic systems: task success metrics, trajectory evaluation, hallucination analysis, and regression detection across the learning flywheel.
Develop offline and online evaluation loops — including LLM-as-judge frameworks — that guide rapid iteration and provide the ground truth signal the flywheel depends on.
Lead systematic experimentation across prompts, agent configurations, model variants, and tool integrations.
Own bounded, end-to-end ML workflows from problem framing through deployment, monitoring, and lifecycle management.
Partner with engineering on integration, observability, and production readiness — without acting as a full-time infrastructure owner.
Identify systemic gaps across the ML stack (accuracy, latency, cost, reliability) and lead the work to close them.
Act as the technical reference point for agentic AI and SLM best practices across the team.
Drive cross-functional alignment with product and engineering through evidence-backed technical recommendations that influence the roadmap.
Mentor senior and mid-level engineers on experimentation methodology, evaluation design, and production ML system development.