Skills

Go Research Training Machine Learning PyTorch Deep Learning Autonomy OpenAI Langchain Large Language Models

Job Specifications

Machine Learning Engineer: LLM Interpretability & Systems

Location: San Francisco, CA - 5 days a week onsite

About Our Client & The Mission

We are partnering with a San Francisco-based AI company pushing open-source Large Language Models beyond what most teams think is possible. They do not just prompt models. They open the black box. By leveraging cutting-edge interpretability techniques (XAI), they manipulate model internals to push open-source models to rival and beat Frontier models in both capability and cost-efficiency.

They are primarily looking for a highly driven Machine Learning Engineer who can take complex, cutting-edge XAI research and translate it into robust, deployable production systems.

The Role

This is not a prompt engineering or RAG-wrapper position. You will be operating deep within the stack, interacting directly with model weights, activations, and architectures. Your mandate is simple but exceptionally difficult: state exactly how a model can be improved for a specific purpose using interpretability, and then build the system to do it.

What You Will Do:

Translate Research to Production: Take and improve the latest papers in mechanistic interpretability, activation engineering, and representation engineering and turn them into scalable production code.
Model Surgery: Apply techniques like control vectors, activation patching, and feature extraction to fundamentally alter and enhance model capabilities.
Maximize Open-Source: Squeeze every ounce of performance out of open-weight models to make them hyper-efficient and highly competitive against closed-source Frontier models.
Build the Pipeline: Design and implement the infrastructure required to deploy these modified, highly-optimized models into production environments.

Who You Are

Deep Technical Fundamentals: You possess a rigorous understanding of Transformer architectures, PyTorch internals, and the mathematical foundations of deep learning.
Beyond the API: You have hands-on experience training, fine-tuning, and dissecting LLMs. You understand the difference between superficial augmentation and structural model improvement.
High Agency & Conviction: You do not wait for tasks to be assigned. You identify bottlenecks, propose deeply technical solutions, and execute them with absolute conviction.
Mission-Obsessed: You are fundamentally aligned with the mission to democratize state-of-the-art AI capabilities through open-source optimization.

The Anti-Persona (Who this role is NOT for)

If your primary experience involves calling the OpenAI API, chaining together prompts with LangChain, or building standard RAG pipelines without touching the underlying model infrastructure, this role is not a fit. We need someone who can go into the matrix, not just interact with the interface.

What We Offer

Aggressive Equity & Compensation: Highly competitive salaries and significant equity stakes to incentivize the hard work required. $180-260K, 0.5-1% equity
Impact: Pioneering work at the intersection of XAI and production engineering, directly shaping the capabilities of the core product.
Autonomy: A high-trust environment where technical conviction is valued and supported.

About the Company

Shields Group Search specializes in serving early stage VC and PE backed companies. We offer recruitment and hiring solutions that are industry-agnostic with a particular emphasis on commercial and operational roles. Leveraging tech enabled sourcing and an unwavering commitment to selectivity, our firm identifies and recruits the best candidates to meet your business needs. Backed by a team of seasoned professionals, we ensure confidentiality, quality, and efficiency, in every search, providing our clients with a distinctive... Know more