Ads Content Understanding (ACU) is Reddit’s core commercial content understanding team for Ads. The team owns and produces signals that describe what Reddit content is about, how brand safe and suitable it is, and what users are trying to accomplish in commercial conversations. ACU is responsible for:
- The Knowledge Graph (entities, brands, products, and relationships across Reddit and external sources).
- Content taxonomies such as IAB, Shopify Standard Product Taxonomy, IAS, and other commercial taxonomies used for targeting, safety, and marketplace dynamics.
- Opinion mining for ads use cases: sentiment, stance, commercial intent, and other qualitative attributes of conversations.
- Shopping / product understanding: detecting product entities, product categories, and product attributes in organic conversations and aligning them with shopping catalogs.
- Signals and tags registry: a unified, governed catalog of ACU signals that powers retrieval, ranking, safety, and insights across Ads Foundations and partner teams.
We are looking for a Staff Machine Learning Engineer who will lead the Commercial Content Understanding roadmap for the Monetization org and act as the technical owner for ACU’s signals and ML systems. Roughly 50% of their time should be spent in technical leadership and mentorship (driving designs, standards, cross-team alignment), and 50% in direct hands-on work (modeling, pipelines, and debugging complex production systems).
Responsibilities:
- Provide technical leadership and mentorship to MLEs and SWEs doing ML work in ACU, acting as de facto tech lead for content understanding and signals: driving design reviews, setting technical standards, and uplifting the team’s modeling and systems craft.
- Develop evaluation systems and quality monitoring systems for content understanding signals, using SOTA LM-judge practices.
- Drive operational excellence for ACU’s ML systems by defining SLOs, alerting, and dashboards for key signals (coverage, latency, precision/recall, cost)
- Build and evolve content understanding capabilities for commercial conversations (e.g., reviews vs. recommendations vs. comparisons vs. Q&A; sentiment and stance; product entities and categories) and operationalize them as robust signals that power contextual and shopping ads, auto-targeting, new formats, and insights products.
- Lead design and implementation of signals pipelines and produce an ACU signals registry. Partner with platform teams and other content understanding teams to ensure efficient, reliable serving at Reddit scale.
- Drive LLM and modern ML best practices within ACU: define when to prompt, finetune, or distill; design evaluation and safety harnesses; and lead at least one major distillation effort to replace external APIs with in-house models.
- Operate across the full ML lifecycle (problem framing, data, modeling, evaluation, deployment, monitoring, and oncall), designing scalable, resilient MLOps pipelines and championing responsible AI (bias, safety, explainability) for ACU’s models and signals in production.
Required Qualifications:
- 7+ years of relevant MLE experience delivering production ML systems (models + pipelines + serving) at scale, ideally in large-scale content understanding domains, or Ads.
- Demonstrated Staff-level technical leadership: has driven architecture decisions, standards, and design reviews across multiple teams, and has aligned PMs, DSs, and engineers on shared ML systems or platforms without direct people-management authority.
- Excellent communication skills, with the ability to explain complex technical trade-offs to PMs, DSs, and other engineering teams, especially in ambiguous, cross-team problem spaces like Seekers/Searchers monetization.
- Strong track record building and shipping NLP / Language models / content understanding models to production (e.g., classifiers, encoders, sequence or session models), with clear business outcomes (e.g., CTR/ROAS uplift, safety improvements). Experience with commercial or intent modeling is a strong plus.
- Practical experience using LLMs in production for labeling, evaluation, or distillation (e.g., LM-as-judge, prompt-based classifiers, LLM-generated labels distilled into smaller models), including managing quality, cost, and latency trade-offs.
- Deep experience with PyTorch, TensorFlow, or similar, and production-quality code in Python (and ideally one statically typed language like Go/Java/C++). Comfortable owning training, evaluation, and deployment code end-to-end.
- Experience designing ML systems and pipelines: offline training, feature pipelines (batch/streaming), online serving, monitoring, and experimentation for high-traffic surfaces.
Benefits:
- Comprehensive Healthcare Benefits and Income Replacement Programs
- 401k with Employer Match
- Global Benefit programs that fit your lifestyle, from workspace to professional development to caregiving support
- Family Planning Support
- Gender-Affirming Care
- Mental Health & Coaching Benefits
- Flexible Vacation & Paid Volunteer Time Off
- Generous Paid Parental Leave
#LI-Remote
