FULL-TIME ONLY, NO CONSULTANTS- ONLY local candidates from NYC, NJ, CT and must interview in-person
- 5 day work week is a must
- Seeking smart, hardworking candidates from top 40 colleges/universities in USA.
- Compensation: $250,000–$315,000 Base Salary + Performance Bonus + RSUs
Job Title: Senior Agentic AI, AI/ML Engineers
Employment Type: Full-Time
Work Location: Midtown East, NEW YORK CITY
Job Description
We are looking to hire multiple top-tier Senior Agentic AI Engineers for a leading technology-driven company that develops advanced AI platforms for traders and portfolio managers .
The ideal candidates will have strong hands-on engineering experience, particularly in Python programming, Artificial Intelligence, and Machine Learning . ectively.
Must-Have Requirements
- Strong hands-on Python programming experience – Mandatory.
- Extensive experience in Agentic AI, Artificial Intelligence and Machine Learning
- Combination of skills below:
1. Advanced LLM Orchestration & Frameworks
- Agent Framework Mastery: Deep experience with frameworks designed for multi-step reasoning and loops, such as , , AutoGen, or CrewAI.
- Prompt Engineering for Autonomy: Crafting robust system prompts, few-shot examples, and chain-of-thought (CoT) structures that guide agents to self-correct and debug their own outputs.
- State Management: Managing conversational and execution state across long-running, asynchronous agent loops to prevent context degradation and memory loss.
2. Planning, Reasoning & Decision-Making Paradigms
- Architectural Patterns: Implementing advanced design patterns like Reflection (self-critique), Tool Use (Function Calling), Planning (breaking complex goals into sub-tasks), and Multi-Agent Collaboration.
- Reasoning Frameworks: Familiarity with approaches like ReAct (Reason + Act) and Tree-of-Thoughts (ToT) to help agents dynamically choose the best next step.
3. Tool Integration & API Engineering
- Robust Tool Use: Designing reliable interfaces for agents to interact with external systems, databases, internal enterprise tools, and web APIs.
- Structured Output Parsing: Utilizing tools like Pydantic or native JSON mode to ensure agents communicate with downstream code reliably without breaking due to malformed LLM responses.
4. Retrieval Systems & Memory Management
- Advanced RAG (Retrieval-Augmented Generation): Implementing semantic search, hybrid search, reranking models, and graph-based knowledge retrieval (GraphRAG) to feed accurate context to agents.
- Vector Databases: Proficiency with vector stores (such as Pinecone, Milvus, Qdrant, or pgvector) for efficient context storage and retrieval.
- Memory Tiers: Architecting short-term working memory (context window management) and long-term memory (episodic and semantic storage) for stateful agents.
5. Evaluation, Safety & Guardrails
- Agent Evaluation: Developing automated evaluation pipelines (using frameworks like Phoenix or TruLens) to measure agent success rates, loop efficiency, and failure modes across thousands of scenarios.
- Guardrails & Security: Implementing safety frameworks (like NeMo Guardrails) to prevent prompt injection, rogue looping, unauthorized tool execution, and toxic generation.
