Applied Researcher

Job description:
The Applied Researcher – AI Security is responsible for conducting applied research in AI security with a focus on analysing, evaluating, and strengthening AI models and pipelines. The role bridges theory and practice by translating research ideas into validated prototypes, experimental results, and reusable security techniques that enable secure, compliant, and trustworthy AI systems.
Key Responsibilities
- Conduct applied research across AI security topics including model behaviour analysis, provenance analysis, and model unlearning.
- Design and execute experiments, benchmarks, and comparative evaluations for AI security techniques.
- Develop proof-of-concept implementations and research artefacts to demonstrate feasibility and effectiveness.
- Analyse model internals such as weights, activations, and outputs to identify anomalous or insecure behaviour.
- Collaborate closely with Research Leads and engineers to operationalise research outcomes.
- Document methodologies, results, and findings for internal reuse and external publication.
- Contribute to technical papers, patents, and research disclosures as required.
Required Qualifications & Experience
Education:
- Master’s or PhD degree (CS, AI/ML, Crypto, Cybersecurity) from Premier Institute** with research credentials in Artificial Intelligence, Machine Learning, Computer Science, Cybersecurity, or a related field.
Experience:
- 5–8+ years of experience in applied AI/ML research or security research.
- Earlier Experience in R&D labs, Research projects, Product companies, AI Projects, Enterprise solutions, AI Engineering
- Publications in top tier conferences (mandatory)
- Demonstrated experience translating experimental ideas into working prototypes.
** It is suggested to prefer Post-Doc, in case difficult to find suitable candidates with 5-8 yrs of relevant experience**
Required Technical Skills
- Strong understanding of large language models and modern ML architectures, including fine-tuning, LoRA/adapters, and RAG systems.
- Understanding of model internals including weight-space behaviour, activation patterns, and interpretability methods.
- Hands-on exposure to AI security techniques such as model fingerprinting, backdoor analysis, derivation detection, and behavioural testing.
- Familiarity with model unlearning concepts, verification approaches, and privacy risk mitigation.
- Proficiency with ML experimentation frameworks (e.g., PyTorch, JAX) and reproducible research workflows.
- Strong programming skills in Python and experience working with research codebases.

Work arrangement
No