SPARC is an AI research and engineering lab focused on building secure, sovereign, and continuously improving AI systems. We operate at the frontier of AI and security, driven by a two-fold mission:

- Self-improving AI: We advance AI systems that improve through experience—learning from interaction, refining their reasoning, and growing more capable over time.
- Secure-by-Design Infrastructure: We develop the foundations for AI infrastructure that is secure by design, verifiable at scale, and resilient enough for deployment in critical sectors such as defence, healthcare, finance, and national infrastructure.

Cybersecurity serves as our proving ground. It provides objective feedback, rapidly evolving environments, and adaptive adversaries, making it one of the most demanding domains for developing autonomous AI systems. Success in this environment pushes the frontier of AI itself, advancing systems that can reason, adapt, and operate reliably in complex real-world settings. At SPARC, you'll work on fundamental research and engineering challenges while building AI systems designed to solve problems of national and strategic importance. This is a place for those who want to move beyond incremental work and contribute to systems that matter.

Roles

As a AI Researcher, you will develop novel algorithms and systems for autonomous AI agents in cybersecurity. The role focuses on advancing foundation models, autonomous AI agents, reasoning systems, reinforcement learning, and continual adaptation for complex real-world environments.

You will collaborate closely with other AI researchers, engineers,



and cybersecurity experts to transform cutting-edge research into practical, high-impact systems. This is a research-focused role with significant opportunities to influence the future of AI agents, long-horizon reasoning and their applications in cybersecurity.

What you will work on

- Post training and Reasoning. Develop novel post-training techniques for reasoning models, including reinforcement learning and synthetic data generation. Improve model capabilities for planning, tool use, long-horizon reasoning, and decision-making. Research continual and experiential learning methods for self-evolving AI systems.
- Agent Systems and Harness Optimization. Design and build autonomous AI agents capable of operating in complex environments. Develop algorithms for agent optimization, including memory systems, verifiers, self-reflection, planning, and tool orchestration. Research methods for improving reliability, robustness, and generalization of AI agents.
- Evaluation abd Benchmarking. Design rigorous evaluation methodologies and benchmarks for reasoning models and AI agents. Build scalable infrastructure for measuring agent performance, reliability, and safety. Develop datasets and environments that enable reproducible research on autonomous systems.




- Research and Impact. Build research artifacts and contribute to the broader AI community. Translate research breakthroughs into deployable AI systems used in real-world cybersecurity settings.

Qualifications We are looking for researchers with deep expertise in one or more of the following areas:

- Large Language Models and Post-Training
- Reinforcement Learning and RL for Reasoning
- Agentic AI (tool use, memory, planning, verification)
- AI Evaluation and Benchmarking
- Machine Learning and Foundation Models
- Probabilistic Reasoning and Decision Making
- Distributed AI Systems and Inference Infrastructure

Candidates should hold a PhD in Computer Science, Artificial Intelligence, Machine Learning, or a related field, or possess equivalent research experience. A strong publication record in leading conferences such as NeurIPS, ICML and ICLR is highly desirable. Why join SPARC?

- Work on some of the hardest open problems in AI agents and reasoning.
- Build AI systems that operate in real-world, high-stakes environments.
- Collaborate with researchers and engineers across AI and cybersecurity.
- Help define the future of autonomous cybersecurity systems through an open research workplace and collaborations.

Ideal candidates are excited by questions such as: How do we train reasoning models that can solve long-horizon tasks?

How do we build self-improving agents with memory and verification?

How do we rigorously evaluate autonomous AI systems?

How do we make AI agents reliable enough for real-world cybersecurity?

📌 Artificial Intelligence Researcher (Bengaluru)
🏢 SPARC (Security, Privacy and AI Research Centre)
📍 Bengaluru