Artificial Intelligence Researcher

About Swirl

Swirl is building the Palantir for Agentic Commerce: a low/no-code framework that empowers enterprises to create multimodal AI agents with observability, governance, and deep customer context baked in. We help OEMs and consumer brands ship agentic AI solutions in daysnot months.

The Mission (Why This Role Exists)

Push the boundaries of what's possible with AI at Swirl. As an AI Researcher, your mission is to explore new frontiers in large language models (LLMs), architectures, and multimodal systemsand translate those insights into proofs of concept (POCs) that can evolve into production-grade features. You'll operate at god speed: researching, testing, and prototyping cutting-edge capabilities that give Swirl and its customers a competitive edge.

What You'll Do

Research & Innovate: Continuously evaluate emerging LLMs, multimodal models, and next-generation interaction paradigmsincluding voice, agentic UI, and adaptive interfacesfor performance, usability, latency, and extensibility across OpenAI, Anthropic, Mistral, DeepSeek, Meta, and others.

Prototype Rapidly: Build high-fidelity POCs using Swirl's framework (LangGraph/LangChain-style patterns, structured outputs, retrieval workflows, tool orchestration).

Architect Experiments: Design experiments to test new agent behaviors, fine-tuning methods, and memory architectures.

Collaborate with Engineering: Partner with AI Engineers to translate research prototypes into production-ready systems.

Benchmark & Evaluate: Develop standardized evaluation harnesses to measure model quality, latency, cost, hallucination rates, and factual consistency.

Stay Ahead of the Curve: Track advancements in LLM architecture, adapters, multimodal I/O, and agent alignment. Publish Internally: Document key findings, present learnings, and propose next-gen features or platform improvements.

You Might Be a Fit If You Have

2+ years of hands-on AI research or engineering experience (academic or industry).

Strong knowledge of LLM internalstransformer architecture, attention mechanisms, fine-tuning, adapters (LoRA/QLoRA), and inference optimization.

Proficiency in at least one modern ML language and framework e.g. Python, PyTorch, JAX, or Hugging Face Transformers.

Experience building or fine-tuning retrieval-augmented and tool-using agent systems.

Working knowledge of system architecture, including cloud compute, GPUs, and model serving pipelines.

Proven ability to rapidly prototypetaking an idea from research paper to working demo within days.

Strong communication skills; able to explain complex research to product and engineering teams.

Nice-to-Haves

Familiarity with LangGraph, LangChain, or DSPy.

Experience in multimodal AI (vision, ASR/TTS, video, speech).

Experience with evaluation frameworks, red teaming, or alignment research.

Contributions to open-source LLM projects or published research.

Background in enterprise AI systems, governance, or compliance-aware architectures.

What Success Looks Like

30 Days: Deliver your first experimental prototype demonstrating a novel agentic capability.

90 Days: Publish 2+ internal research notes and ship a working POC that demonstrates measurable improvement in capability or performance.

6 Months: Establish a repeatable framework for AI experimentation and influence Swirl's platform roadmap through innovation.

Why Swirl

Work directly with the founder and core platform team on defining the frontier of agentic AI.

Ship innovation that becomes the backbone of enterprise-grade AI agents.

Be part of a fast-moving, research-driven team where ideas become production features at record speed.

Reach Out To: [Confidential Information]