--- title: 'ML Research Engineer, Trust & Safety at Realm Labs' canonical: 'https://feeny.ai/job/ml-research-engineer-trust-safety-realm-labs-sunnyvale-js9kd2qfaq4s' type: 'job' last_seen: '2026-11-03' --- # ML Research Engineer, Trust & Safety at Realm Labs - **Company:** Realm Labs - **Location:** Sunnyvale, CA - **Compensation:** $200k–$260k - **Employment:** full-time - **Work type:** onsite - **Posted:** 2026-01-16 - **Last confirmed live:** 2026-11-03 - **Apply:** https://jobs.gem.com/realm-labs/am9icG9zdDpaG4cmxb0ORSTneoQfGKnK ## Job description ## Role Overview This is an individual contributor role blending ML research, applied ML, and software engineering. Topics include trust and safety challenges for large language models (LLMs) and multimodal LLMs (MM-LLMs): - Prompt injection attacks and adversarial robustness - Safety alignment and guardrails - Privacy and confidentiality - (Mechanistic) interpretability ## Responsibilities - Designing and executing research experiments on LLM trust and safety - Reading, synthesizing, and producing research papers and technical write-ups - Building, running, and owning systems end-to-end - Contributing to core product capabilities as a founding team member ## What We're Looking For ML & Research - Demonstrated research experience; PhD in a technical field strongly preferred - Ability to identify relevant research questions and find answers through literature review or experimentation - First-author or major contributing authorship on peer-reviewed publications; please list these on your resume or cover letter - Hands-on experience with PyTorch, HuggingFace (Transformers, Datasets), and applied deep learning - Experience training and evaluating deep learning models - Nice to have: LLM fine-tuning, multimodal LLMs, PEFT/LoRA - Nice to have: Familiarity with ML interpretability methods (mechanistic interpretability, sparse autoencoders, linear probes, NLP/Vision interpretability) Software Engineering - Proficiency in Python; experience with Jupyter notebooks - Unix environments, Git, and basic AWS/GCP usage - Nice to have: Programming languages well-roundedness; experience with statically-typed and functional programming languages - Nice to have: Familiarity with (LLM) deployment tooling like Docker, vLLM, Triton Inference Server or similar Logistics - Available to start within 2 months - Role may include limited on-call responsibilities tied to production ownership Additional Information - This is a founding, high-ownership role with direct impact on core product capabilities. - You will be expected to build, run, and own systems end-to-end. - The role may include limited on-call responsibilities aligned with production ownership. ## About RealmLabs - AI systems misbehave: they leak sensitive data, get manipulated through prompt injections, or behave in ways their builders never intended. RealmLabs is building the infrastructure to detect, debug, and prevent these behaviors. - Our approach is to secure AI from within: observing models from the inside, extracting signals relevant to their behavior and misbehavior, and patching them on the inside, rather than relying on input/output filters or external guardrails. - We are a pre-seed startup and a finalist in the RSAC 2026 Innovation Sandbox. Our customers include Anthropic, a leading ride-sharing platform, and a Big 3 management consulting firm. - This is a founding engineer role with meaningful equity at an early-stage company. ## Compensation & Benefits - Market aligned compensation and benefits. - Founding equity (Equity is a significant component of this role and will be discussed) - Medical, Dental, Vision, Life insurance, 401-K, In-office lunch etc. Visa sponsorship - We do sponsor visas! However, we aren't able to successfully sponsor visas for every role and candidate. But if we make you an offer, we will make all reasonable effort to get you a visa, and we retain an immigration lawyer to help with this. ## About Realm Labs ## Company Overview - **One-liner**: Realm Labs provides enterprise AI security and monitoring solutions that detect and prevent failures in LLM applications using deep neural inspection. - **Entity Type**: Private (early stage startup) - **Headquarters**: United States (specific city not disclosed) - **Founded**: 2023 - **Founders**: Saurabh Shintre (CEO & Founder), Akash Mukherjee (Head of Engineering), Piotr Mardziel (Head of AI) ## Core Business - Primary industry: AI Security / Cybersecurity - Target customers: B2B Enterprise (companies deploying large language models in production) - Mission: “We exist to ensure that progress doesn’t come at the cost of control, clarity, or accountability.” ## Products & Services - **Realm Labs Platform**: AI-native SaaS/API that monitors every LLM interaction in real time. Uses patented Deep Neural Inspection (DNI) to read a model’s internal representations and catch hallucinations, prompt injections, jailbreaks, data leakage, and unsafe behavior at the source. Three modules: *Detect* (surface failures), *Respond* (block, rewrite, or route problematic outputs), *Improve* (feedback loop to correct drift and prevent repeat incidents). ## Market Standing - **Valuation**: Not disclosed (no reported funding rounds) - **Key Metric**: Headcount ~7 employees (as of late 2025/early 2026) - **Notable Investors/Partners**: Not publicly listed - **Growth Signals**: Early-stage research-driven company with a clear product and a team holding 20+ patents and 5,000+ academic citations across adversarial ML, AI safety, and system security. ## Competitive Advantages - **Deep Neural Inspection** – only inspection point inside an LLM’s hidden representations that the model itself cannot forge, unlike surface-level output analysis. - Founding team brings decades of combined expertise from Symantec, Splunk, Google, Apple, Snowflake, and leading universities (CMU, Stanford, USC). - Focus on both detection and automated response creates a closed-loop security posture for enterprise AI. ## Strategic Focus - Current priorities: Building trust and accountability in AI systems by providing guardrails, monitoring, and incident response. Growth direction: expanding from LLM security to a broader AI security platform that covers multi-modal models and agentic workflows. ## Why Work Here - **Culture**: Small, agile, research-first environment where engineers and researchers solve hard AI safety problems. - **Work Model**: Hybrid (remote + on-site in the US) based on Built In profile. - **Notable Highlights**: Opportunity to work at the intersection of cutting-edge AI and security, influence product direction from day one, and collaborate with founders who are recognised experts in the field. ## Sources 1. [builtin.com](https://builtin.com/company/realm-labs) – Company overview, headcount, hybrid work, founding year 2. [realmlabs.ai/company](https://www.realmlabs.ai/company) – Founding team, mission, expertise 3. [realmlabs.ai](https://www.realmlabs.ai/) – Product details (Detect, Respond, Improve) and Deep Neural Inspection ## Other roles at Realm Labs - [Machine Learning Research Intern (Winter/Spring 2027)](https://feeny.ai/job/machine-learning-research-intern-winter-spring-2027-realm-labs-sunnyvale-qfv0hsht8cgm) — Sunnyvale, CA - [Intern: AI Red Teaming (Fall 2026)](https://feeny.ai/job/intern-ai-red-teaming-fall-2026-realm-labs-sunnyvale-cc16r5a6pk2v) — Sunnyvale, CA - [Marketing Lead](https://feeny.ai/job/marketing-lead-realm-labs-united-states-7kbp1qe6tzxp) — United States - [Software Engineer, ML Infrastructure](https://feeny.ai/job/software-engineer-ml-infrastructure-realm-labs-sunnyvale-820mmna8cyxn) — Sunnyvale, CA