Hire by role
Hire vetted AI Security Engineers and AI Red Teamers who secure LLM applications and agents. Shortlist in 48 hours; working within 4 working days.
Fig. 01Cybersecurity
AI Security Engineer
Cybersecurity
AvailabilityShortlist in 48 h
Vetted AI Security Engineer profiles free to start within 4 working days of signature. Counts appear in bands (5–10, 10–20, 20+); below 5 a role shows “Shortlist in 48 h” instead of a number.
Indicative monthly rate per seat
| Level | Experience | From (USD / month) |
|---|---|---|
| Mid | 4–6 years | 5,120 |
| Senior | 6–9 years | 6,730 |
| Lead | 9+ years | 8,740 |
Quotation-based. VAT shown separately. View in SAR, AED, USD or EUR. Quotes valid for 30 days.
Includes the Cybersecurity practice rate of +15% on the base band.
Remote Pro Ai provides vetted AI Security Engineers and AI Red Teamers who secure LLM applications and agents. They red-team apps for prompt injection, jailbreaks and data exfiltration with tools such as garak, PyRIT and Promptfoo, design guardrails and tool-permission boundaries, and build AI security tests into CI/CD.
Your AI Security Engineer red-teams LLM apps for prompt injection, jailbreaks and data exfiltration, and assesses model supply chain, training-data and fine-tuning risks. Typical stack: garak, PyRIT, Promptfoo, NeMo Guardrails and Lakera.
AI security is the newest premium specialism in cybersecurity. Enterprises are connecting copilots and agents to internal data and tools faster than security teams can assess them, and the OWASP LLM Top 10 and ISO/IEC 42001 now give auditors something to test against. Very few practitioners have hands-on AI red-teaming experience.
Illustrative profiles that mirror the vetted bench for this role. Names, employers, full gate scores and video introductions, where recorded, are shared after the click-through NDA.
AI Security Engineer
Senior8 years
Remote Pro Ai Vetted Verify credential RP-5485
AI Security Engineer
Lead10 years
Remote Pro Ai Vetted Verify credential RP-3866
AI Security Engineer
Mid6 years
Remote Pro Ai Vetted Verify credential RP-2247
Illustrative profiles that mirror the vetted bench for this role, anonymised. Names, employers, full scores and credentials are shared after the mutual NDA in the booking flow.
Red-teams LLM apps for prompt injection, jailbreaks and data exfiltration
Designs guardrails, output filtering and tool-permission boundaries
Assesses model supply chain, training-data and fine-tuning risks
Builds AI security testing into CI/CD
Maps controls to OWASP LLM Top 10 and ISO 42001
Toolset on our bench
We test each capability in a live technical assessment scored by a practice lead. Nothing on this list is taken from a CV.
Mid4–6 years
Delivers well-scoped work independently within your architecture and standards.
from USD 5,120/month
Senior6–9 years
Owns a component or workstream end to end, reviews code and mentors.
from USD 6,730/month
Lead9+ years
Sets technical direction, leads the squad and reports to your leadership.
from USD 8,740/month
We never place junior profiles. A practice lead confirms every level, so the band you buy is the scope you get.
No single certification defines this role, so we test the work itself: a live technical assessment scored by a practice lead, plus references from shipped projects.
Remote Pro Ai information security: ISO/IEC 27001 certification in progress (target Q2 2027) · SOC 2 Type II planned for 2027. Evidence and policies are in the Trust Center.
Not sure this is the right role? Describe the outcome you need and the Scope Agent maps it to roles, levels and a draft squad in minutes.
A Saudi insurer
Arabic GenAI knowledge assistant
6 engineers · 7 months
Bilingual RAG assistant with evaluation harness, human-in-the-loop review, monitoring, one-click rollback and PDPL-compliant logging
MatchIQ™ is our vetting and matching engine: 3.4% of applicants pass 5 gates. The live assessment for this role tests the 5 capabilities above against a realistic task, not trivia.
Self-serve bookings receive a shortlist within 48 hours and enterprise requests get named candidates within 72 hours. After you approve a profile and sign, your specialist is working within 4 working days, whether the request is on the bench, off-bench or on-site. When fewer than 5 AI Security Engineer profiles are on the bench, we shortlist from the wider vetted network in the same 48 hours.
Pricing is quotation-based and depends on level, hours and any on-site, clearance or data-residency needs. Current bands: Mid from USD 5,120/month · Senior from USD 6,730/month · Lead from USD 8,740/month. VAT is shown separately, quotes can be viewed in SAR, AED, USD or EUR, and every quotation is valid for 30 days. The Squad Builder shows your quote instantly.
Remote Pro Ai is contractually responsible for quality. Raise it at any time: a replacement starts within 4 working days; the full handover completes in under 14 days. Your first engagement also begins with a no-bill trial, one per company: 5 working days on self-serve bookings and 10 on enterprise contracts.
This role is offered at Mid, Senior and Lead level. We never place junior profiles. A practice lead confirms each specialist’s level through a live technical assessment and reference checks, so the band on your quotation reflects scope the person has already delivered rather than a job title on a CV.
Engineers attack your LLM application the way an adversary would: direct and indirect prompt injection, jailbreaks, system-prompt extraction, data exfiltration through retrieval or tools, and unsafe agent actions. You receive reproducible findings rated by impact, each with guardrail and design fixes.
garak, PyRIT and Promptfoo for automated adversarial testing, NeMo Guardrails and Lakera for runtime protection, and the OWASP LLM Top 10 and ISO/IEC 42001 for mapping controls. Automated scans are combined with manual testing, because the most damaging attacks are rarely in any scanner’s library.
Yes. Engineers build adversarial suites into CI/CD, so every prompt, model or retrieval change is tested before release. Results are tracked over time, which matters because model upgrades can silently reopen vulnerabilities you have already fixed. Failing tests block a release in the same way a failing unit test would.
Yes. They review where models, datasets and adapters come from, check for poisoned or untrusted training data, assess fine-tuning pipelines and evaluate third-party providers’ controls. The result is a supply-chain risk view your governance and procurement teams can act on.
Add this role to a squad, see your quote instantly, and hold profiles free for 72 hours.
Build your squad online
No card needed · 72 h free hold · No-bill trial: 5 working days self-serve, 10 in the Enterprise lane.
20 minutes with a solutions architect
No sales script.