Handshake currently has 68 active AI-related job listings. The majority of these roles, 60%, are focused on data, with a further 15% in agents. Research and Engineering are the most frequent functions hiring for these positions. Recent hiring activity shows a significant increase, with 23 new AI roles posted in the last 30 days, representing a 92% rise compared to the preceding 30-day period.
Currently tracking 23 active AI roles, down 48% versus the prior 4 weeks. Primary focus: Agent · Engineering.
Handshake currently has 65 active AI-related roles in our index. The most common open titles are: Music Producer - AI Trainer (2), Strategic Projects Lead, Coding (2), 3D Slicer Specialist - AI Trainer , AI Red Teamer, LLM Generalist, Analog Engineer - AI Trainer. Most positions are in Engineering and Research.
Handshake's active AI hiring is concentrated in: data (66%), agents (15%), application (8%). These categories follow a seven-stage AI lifecycle: data, pre-training, post-training, serving infrastructure, agents, evaluation, and application.
Handshake is hiring AI talent in: United States (22 roles), India (4 roles).
Job postings at Handshake most frequently reference: evals, synthetic data, model serving, agent orchestration, llm observability.
In the past 30 days, Handshake has posted 10 new AI-related roles. That is a -57% change versus the prior 30 days (23 → 10).
| Title | Stage | AI score |
|---|---|---|
| AI Red Teamer, CBRNE This role focuses on evaluating AI models for safety and security, specifically concerning CBRNE threats. The Red Teamer will design adversarial prompts, assess model outputs for dangerous knowledge gaps, and document findings to help labs improve model defenses before they reach the real world. This requires deep domain expertise in CBRNE fields, strong ethical judgment, and the ability to think like a threat actor within a structured evaluation framework. | Eval Gate | 9 |
| AI PhD Student Researcher - Fall 2026 Handshake AI is seeking a PhD Student Researcher to work on novel RLHF/GRPO pipelines, instruction-following refinements, reasoning-trace supervision, multilingual/long-horizon/domain-specific benchmarks, automatic vs. human preference studies, robustness diagnostics, active-learning loops, data value estimation, synthetic data generation, and low-resource fine-tuning strategies. The goal is to produce an archive-ready manuscript or top-tier conference submission. |
| Post-trainEval Gate |
| 9 |