Upwork Inc.’s (Nasdaq: UPWK) family of companies connects businesses with global, AI-enabled talent across every contingent work type including freelance, fractional, and payrolled. This portfolio includes the Upwork Marketplace, which connects businesses with on-demand access to highly skilled talent across the globe, and Lifted, which provides a purpose-built solution for enterprise organizations to source, contract, manage, and pay talent across the full spectrum of contingent work. From Fortune 100 enterprises to entrepreneurs, businesses rely on Upwork Inc. to find and hire expert talent, leverage AI-powered work solutions, and drive business transformation. With access to professionals spanning more than 10,000 skills across AI & machine learning, software development, sales & marketing, customer support, finance & accounting, and more, the Upwork family of companies enables businesses of all sizes to scale, innovate, and transform their workforces for the age of AI and beyond.

Since its founding, Upwork Inc. has facilitated more than $30 billion in total transactions and services as it fulfills its purpose to create opportunity in every era of work. Learn more about the Upwork Marketplace at Upwork.com

We’re looking for a Lead Machine Learning Engineer/Scientist to help build a Dynamic Memory Management capability for our LLM-powered experiences (including agentic systems and tool-using assistants). This role sits at the intersection of retrieval, memory, reasoning, and orchestration - designing how AI systems store, update, compress, retrieve, and apply knowledge across sessions, tasks, and workflows.

In this role, you’ll focus on building production-grade memory architectures that integrate structured and unstructured signals, including user preferences, entities, constraints, conversation history, tool results, marketplace context, and long-term facts. You’ll design “memory policies” (what to store, when to forget, how to summarize), develop RAG + memory fusion strategies, and train or post-train models to reliably execute function calls/tool calls grounded in memory and context.

You’ll partner closely with engineering, research, product, and trust & safety teams to transition memory research prototypes into robust, measurable, production-ready systems, improving personalization, reducing hallucinations, increasing task success rates, and enabling more capable autonomous workflows.

Responsibilities

• Design and build a Dynamic Memory Management system for LLM/agent applications, including memory ingestion, CRUD operations, retrieval, consolidation, summarization, and forgetting policies.
• Develop RAG + memory architectures that integrate vector databases, relational databases, and knowledge graph-based representations for context-aware reasoning.
• Create ranking and retrieval strategies (e.g., multi-stage retrieval, re-ranking, memory salience scoring, recency vs. importance tradeoffs, conflict resolution, deduplication).
• Build and evaluate pipelines and models (retrieval-augmented context/knowledge flows), including modeling for memory selection and grounding.
• Train, fine-tune, and post-train models for reliable function calling / tool calling, including:
• tool selection, schema adherence, multi-step tool plans
• post-training with preference optimization / RL-style methods or constrained decoding approaches
• safety-aware and policy-compliant tool use
• Develop or leverage discriminator/verifier models (or LLM-as-judge frameworks) to assess memory correctness, grounding quality, hallucination risk, and tool-call validity.
• Establish offline + online evaluation: memory precision/recall, factual consistency, task success rate, latency/cost, and long-term personalization impact.
• Lead cross-functional efforts to ship memory capabilities end-to-end: data pipelines, privacy boundaries, storage, retrieval, orchestration, monitoring, and iteration loops.
• Mentor engineers/researchers, conduct detailed code reviews, and set best practices for building scalable LLM + retrieval systems.
• Deliver high-quality, measurable outcomes aligned with team and organizational goals.

What it takes to catch our eye

• Proven track record building and deploying LLM-powered agent systems in production with measurable business impact.
• Strong experience with RAG systems, including retrieval pipelines, embedding strategies, re-ranking, and evaluation methodologies.
• Deep practical understanding of memory in LLM systems: long/short-term memory, summarization, consolidation, forgetting, conflict resolution, and personalization.
• Hands-on experience with function calling/tool calling systems and post-training models to reliably invoke tools in structured formats.
• Experience with fine-tuning and post-fine-tuning (e.g., supervised fine-tuning, preference optimization, RL-style approaches), including dataset construction and training pipelines.
• Familiarity with discriminator/verifier models, reward modeling, or automated evaluation strategies for grounding and tool correctness.
• Strong foundations in ranking systems, retrieval modeling, and/or representation learning (dense + sparse + hybrid retrieval).
• Comfortable operating in ambiguity: can define the problem, design experiments, ship incrementally, and improve iteratively with rigorous measurement.
• Excellent software engineering skills: clean code, strong testing discipline, scalable systems mindset.

Come change how the world works.