Research Engineer - Agent Memory

About the role

ABOUT MEM0

We're building the memory layer for AI agents: long-term memory that lets AI remember conversations, learn from interactions, and build real context over time, the difference between a chatbot and something that actually knows you. We're already powering millions of AI interactions for enterprise customers and a fast-growing open-source community (150k+ developers and counting). We're backed by top-tier investors through a $24M Series A & Seed, early enough that the systems you build now become the foundation everything else is built on.

WHY THIS ROLE

Every AI agent today forgets you the moment the session ends. We're fixing that, and you'd be one of the people building the models that decide what an AI remembers, forgets, and learns over time. This isn't research bolted onto a product team: you'll take an idea from a paper on Monday to an A/B test in production within weeks, with millions of real interactions to learn from. If you want your fine-tuning runs to ship, not sit in a notebook, this is that role.

You'll own memory features end-to-end: fine-tune models for extraction, updates, consolidation/forgetting, and dreaming; turn real customer pain points into research hypotheses; implement and benchmark ideas straight out of the literature; and ship them with Engineering at SOTA latency, reliability, and cost. You'll build evaluation at scale, offline metrics and online A/Bs, and close the loop with production feedback to keep pushing quality forward.

This role reports directly to our CTO and is based in-office in San Francisco.

WHAT YOU'LL DO

- Train models: Memory extraction, updates, consolidation/forgetting, and conflict resolution, iterating fast on real data and real outcomes.

- Live at the research frontier: Turn paper ideas into working prototypes in days, benchmark them honestly against baselines, and productionize the ones that actually win.

- Build the evaluation layer: Automated relevance/accuracy/consistency metrics, gold sets, online A/B and interleaving tests, and dashboards people actually check.

- Get close to the problem: Uncover what's actually breaking for customers, turn it into a testable hypothesis, and validate it through real field trials.

- Ship it for real: Design APIs and data contracts, plan safe rollouts, and hit SOTA latency, reliability, and cost, with Engineering, not over the wall from them.

WHAT YOU'LL BRING

- Experience in RAG or information retrieval (retrieval, ranking, query understanding) for real products.

- Model training/fine-tuning experience (LLMs/encoders) with a strong footing in experimental design and iteration.

- Strong Python; deep experience with PyTorch and familiarity with vLLM and modern serving frameworks.

- Built evaluation for complex language and/or retrieval and generation tasks (gold sets, offline metrics, online tests).

- Able to orchestrate data pipelines to run these models in production with low-latency SLAs (batch + streaming).