About BayRock Labs
At BayRock Labs, we pioneer innovative tech solutions that drive business transformation. As a leading product engineering firm based in Silicon Valley, we provide full-cycle product development, leveraging cutting-edge technologies in AI, ML, and data analytics. Our collaborative, inclusive culture fosters professional growth and work-life balance. Join us to work on ground-breaking projects and be part of a team that values excellence, integrity, and innovation. Together, let's redefine what's possible in technology.
Generative AI Developer (LLM Finetuning & Multi-Agent Systems)
We are seeking a hands-on Generative AI Developer to be the primary builder of our production-grade multi-agent applications. This critical role is responsible for implementing complex, stateful workflows using LangGraph, integrating them deeply with Snowflake Cortex AI and Snowpark, and ensuring all agents are highly performant, reliable, and compliant through rigorous testing, observability, and strategic LLM fine-tuning and agent training.
Key Responsibilities
A. Multi-Agent Development & Orchestration
- Agent Implementation: Design, build, and deploy specialized AI agents (e.g., Data Agent, Validation Agent, Assignment Agent) using Python and best practices for modular, re-usable code.
- LangGraph Mastery: Implement complex, long-running, and conditional Multi-Agent workflows using LangGraph, handling state management, human-in-the-loop steps, and robust error handling across critical business processes.
- Prompt & Reasoning: Develop and optimize production-ready prompt templates, manage context windows, and define tool-calling schemas to enhance the agents' decision-making and reasoning capabilities.
B. LLM & Agent Optimization (Finetuning & Training)
- LLM Finetuning: Own the process of finetuning open-source and proprietary foundation models (e.g., via Cortex Fine-Tuning or external platforms) for specific domain tasks (e.g., structured data extraction, classification, complex reasoning) to improve agent accuracy and reduce inference costs.
- Agent Training & Adaptation: Implement strategies to systematically train and adapt agent behavior based on real-world workflow data, focusing on improving tool-use, reasoning chains, and decision-making accuracy within the LangGraph framework.
- Data Curation: Collaborate with data science and engineering teams to curate high-quality, labeled datasets necessary for both pre-training and reinforcement learning techniques for agent improvement.
C. Tooling and Snowflake Cortex Integration
- Snowflake Cortex Integration: Develop custom Agent Tools that interface directly with the Snowflake data layer, specifically leveraging:
- Cortex LLM Functions (CORTEX.COMPLETE) for flexible reasoning tasks.
- Cortex Analyst to execute optimized Text-to-SQL queries for data retrieval and reporting.
- RAG Implementation: Build and optimize the RAG pipeline that allows agents to securely retrieve contextual information (e.g., policy documents, historical contract terms) from Snowflake to ground their responses.
D. Evaluation, Observability & Deployment
- Evaluation Frameworks: Implement systematic testing and evaluation using the LangSmith ecosystem to measure agent performance metrics such as accuracy, groundedness, latency, and cost, tracking improvements post-finetuning and training.
- AI Observability: Integrate logging, tracing, and analytics across the entire LangGraph workflow to provide the auditability and transparency necessary for a critical enterprise application.
- Production Readiness: Assist the architecture team in preparing agents for deployment, including containerization (e.g., Docker) and integration into the production environment (e.g., Snowflake Container Services).
Required Skills & Qualifications
- 6+ years of professional software development experience, with a focus on Python in a data or AI context.
- Hands-on experience building, testing, and productionizing Generative AI applications.
- Expert-level proficiency with LangChain and LangGraph for building complex, stateful multi-agent systems.
- Demonstrated ability to build custom tools and integrate them with Snowflake Cortex AI and Snowpark.
- Strong practical experience in LLM finetuning, including understanding of data preparation, popular techniques (e.g., LoRA), and evaluation of finetuned models.
- Strong practical understanding of RAG (Retrieval-Augmented Generation), semantic search, vector databases, and managing LLM context/memory.
- Experience with evaluation and observability tools for Gen AI (e.g., LangSmith, TruLens, Weights & Biases).
- Familiarity with containerization (Docker) and deployment patterns for AI services.
Pay rate: 25LPA
