Company Description: Curvia AI is a technology and AI services company that combines years of enterprise and Oracle solution expertise with a strong focus on responsible, secure innovation. Built on the heritage of MillionLogics, the organisation is grounded in integrity, innovation, accountability, and a people-first mindset. Curvia AI partners closely with clients to understand their real-world challenges and delivers secure, scalable solutions that turn complexity into chance.

The company emphasises purposeful technology, centring ethics, security, and human values as it explores the potential of artificial intelligence. Team members join a culture that values trust, thoughtful problem-solving, and building a more adaptable and connected future.

About the Role

We are looking for a CVE Security Task Developer. Your job is to take multi-chain CVEs, create offensive and defensive tasks, and package them so current LLM models fail without the solution, while only the client-accepted model can solve them. The final package format will be shared by the client. You should already understand structures like Harbour and proof.json at a basic level.

Offer details:

Mode of work: Fully Remote

Employment: C2H (Contract 2 Hire)

Pay: Competitive

Experience: 1-3 years

Availability: Immediate

What You Will Do

CVE Research

- Review the latest multi-chain CVEs
- Identify the main category (Authentication, Authorisation, Command Execution, etc.)
- Find related CVEs for the same component or service

Offensive Tasks

- Create clear exploit steps for each multi-chain CVE
- Break the attack path into ordered, testable steps

Defensive Tasks

- Create matching defensive tasks for the same CVEs
- Cover detection,



monitoring, mitigation, and hardening

Package & Model Testing

- Build task packages in the structure shared by the client
- Have basic knowledge of formats like Harbour and proof.json (these are examples; final format may differ)
- Make sure the package fails on the latest models like GPT, Gemini, and Claude
- Make sure it passes only on the one model accepted by the client
- Ensure no model can solve the task without the provided solution file
- Test and refine tasks using Cursor and current LLM models

Education

- B.Tech in CSE, AI & ML, or similar engineering degree
- Good base in Data Structures, Algorithms, Operating Systems, and Computer Networks

Skills Needed

- 1–3+ years in cybersecurity and/or AI
- Strong Python skills
- Understand CVEs, multi-chain vulnerabilities, and common vulnerability types
- Can write offensive exploit-step tasks from CVE research
- Can write matching defensive tasks
- Basic knowledge of task package structures like Harbor / proof.json (or similar formats)
- Hands-on experience testing tasks against LLMs (GPT, Gemini, Claude, etc.)
- Comfortable working in Cursor
- Familiar with Prompt Engineering, ML, and Reinforcement Learning
- Experience with Git and CI/CD
- Clear writing and strong problem-solving skills

Nice to Have

- Crowdsourcing / data-labelling experience
- Experience building LLM evaluation or security benchmark tasks
- Certifications like CEH, OSCP, CompTIA Security+, or similar

What We Offer

- Work on real CVE-based offensive and defensive AI security tasks
- Supportive and remote-friendly team
- Pay based on offensive and defensive tasks accepted by our customers
- Growth through onboarding

📌 CVE Security Task Developer (India)
🏢 Curvia AI
📍 India