New opportunity

Principal Scientist

Wipro Limited · Bengaluru, India

About this role

The Principal Scientist – Generative AI & NLP will lead research and development efforts at Wipro's AI Technology Innovation Centre (TIC) in Bengaluru. This role focuses on advancing frontier AI technologies, specifically Large Language Models (LLMs), multimodal AI, and agentic systems, with the goal of achieving on-premise AGI and agentic systems characterized by emotional empathy, self-learning, and multimodal memory. Responsibilities include defining the research roadmap, leading high-impact publications in top-tier conferences (e.g., NeurIPS, ICML, ACL, CVPR), and driving the development of scalable, production-ready AI solutions. The role involves designing novel algorithms, optimizing training pipelines, and architecting end-to-end GenAI systems including RAG and agentic workflows. The successful candidate will mentor junior scientists, collaborate cross-functionally to align research with business goals, and translate discoveries into tangible commercial value. Required qualifications include a Ph.D. or Master’s degree in Computer Science, Machine Learning, Mathematics, Computational Linguistics, or a related field, along with 10–15 years of industry or academic experience in AI/ML, including at least 7 years specifically in NLP and Generative AI. Candidates must demonstrate a strong publication record, deep algorithmic expertise, and hands-on experience in large-scale model training, optimization, and deployment. Proficiency in Python and frameworks such as PyTorch, TensorFlow, and Hugging Face is mandatory, alongside experience with MLOps/LLMOps tools and cloud platforms like AWS, GCP, or Azure.

For full details about this role, click “Apply on company site” to view the employer’s original posting.

Skills for this role

Generative AINLPLarge Language ModelsMultimodal AIAgentic SystemsPythonHugging Face TransformersTensorFlowPyTorchKerasJAXEmbedding modelsVector databasesMLOpsLLMOpsMLflowWeights & BiasesDVCSageMakerVertex AIDistributed systemsData pipelinesAPI development

Checking your profile…

Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.

Understand the role

What matters most

Research Roadmap for On-Premise AGI
Define and execute a strategic research roadmap focused on achieving on-premise AGI and agentic systems. This requires balancing theoretical breakthroughs in emotional empathy, self-learning, and multimodal memory with the practical constraints of on-premise deployment.
Bridging Research and Commercialization
Translate frontier AI discoveries into scalable, production-ready solutions. Success depends on the ability to collaborate with engineering teams to move algorithms from experimental prototypes to commercial products while maintaining scientific rigor.
High-Impact Scientific Output
Lead the publication of research in top-tier conferences such as NeurIPS, ICML, ACL, and CVPR. This priority requires maintaining a high standard of reproducibility and innovation to establish the AI Technology Innovation Centre as a globally recognized leader.
Technical Leadership and Mentorship
Mentor junior scientists and engineers while fostering a culture of high-impact research. This involves guiding team members in best practices for LLM training, optimization, and the development of robust agentic workflows.

Likely challenges to think through

Balancing Frontier Research with System Constraints
Developing novel algorithms for emotional empathy and self-learning often requires massive compute, yet the role mandates on-premise deployment. Prepare to discuss strategies for model optimization, quantization, and efficient inference that do not sacrifice the performance of complex agentic systems.
Cross-Functional Alignment in a Research-to-Product Model
The TIC operates on a research-to-product model, which can create friction between long-term research goals and immediate business delivery timelines. Prepare to demonstrate how you manage stakeholder expectations and ensure that research outputs remain relevant to commercial needs.
Defining Evaluation Metrics for Subjective AI
Building systems with emotional empathy and multimodal memory lacks standardized benchmarks. You will face the challenge of creating original evaluation science to prove the effectiveness and safety of these systems before they reach production.

Practice your answers

Technical practice

Architecting On-Premise Agentic Systems

What to coverDiscuss the challenges of deploying agentic systems in on-premise environments compared to cloud-native deployments. Focus on hardware constraints, data privacy, and the architectural requirements for maintaining multimodal memory and self-learning capabilities locally.

Optimizing Large-Scale Training Pipelines

What to coverExplain your approach to optimizing distributed training for LLMs under compute constraints. Address strategies for model parallelism, data sharding, and efficient checkpointing to ensure reproducibility and performance.

Evaluating Empathy and Reasoning in AI

What to coverHow would you design a robust evaluation framework for subjective AI traits like emotional empathy and long-term memory? Describe the metrics or benchmarks you would use to quantify these capabilities beyond standard loss functions.

RAG and Agentic Workflow Integration

What to coverDescribe the design patterns you use for integrating Retrieval-Augmented Generation (RAG) with agentic workflows. Focus on how you handle prompt orchestration, tool use, and the trade-offs between latency and accuracy in production systems.

Multimodal Model Development

What to coverDiscuss the technical hurdles in aligning different modalities (e.g., text, image, audio) within a single generative model. Explain how you manage embedding spaces and cross-modal attention mechanisms to ensure coherent output.

LLMOps and Lifecycle Management

What to coverDetail your experience with LLMOps tools like MLflow or Weights & Biases for tracking experiments and model versions. Explain how you ensure model lineage and reproducibility when moving from research prototypes to production-ready deployments.

Inference Optimization Techniques

What to coverWhat techniques do you employ to optimize inference for large models, such as quantization, pruning, or knowledge distillation? Explain how these choices impact the balance between model fidelity and real-time performance requirements.

Algorithm Design for Self-Learning

What to coverPropose a methodology for implementing self-learning capabilities in an agentic system. Address how you would manage feedback loops, data drift, and safety guardrails to prevent model degradation during continuous learning.

Behavioural practice

Translating Research to Commercial Value

What to coverDescribe a situation where you had to bridge the gap between a high-level research hypothesis and a tangible commercial product. How did you align your technical roadmap with business stakeholders to ensure the project delivered measurable value?

Mentoring and Research Leadership

What to coverProvide an example of how you have mentored junior scientists to improve their research output or technical skills. Focus on how you balanced their professional growth with the need to meet strict project deadlines.

Navigating Cross-Functional Collaboration

What to coverTell us about a time you worked with engineering and product teams to integrate a complex AI solution. How did you manage conflicting priorities or technical disagreements to ensure a successful deployment?

Driving High-Impact Publications

What to coverDescribe your process for leading a team to produce top-tier research publications. How do you foster an environment that encourages breakthrough thinking while maintaining the rigor required for conferences like NeurIPS or ICML?

Sharpen your knowledge

Topics to prepare

Agentic Systems and Multimodal Memory
The role focuses on building agentic systems with self-learning and multimodal memory. Review current literature on autonomous agent architectures, memory augmentation (long-term vs. short-term), and planning mechanisms. Exercise: Draft a conceptual framework for an agent capable of maintaining state across multimodal inputs, specifically addressing how to evaluate 'emotional empathy' in model responses.
On-Premise AGI and Model Optimization
The position emphasizes on-premise deployment of large-scale models. Study techniques for model compression, quantization, and efficient inference to fit high-parameter models into constrained environments. Exercise: Compare the trade-offs between distillation, pruning, and quantization for a specific LLM architecture, focusing on maintaining performance benchmarks while reducing compute requirements.

Skills to revise

Large-Scale Distributed Training
The role requires hands-on experience in training large models. Practice configuring distributed training jobs using PyTorch or JAX. Exercise: Design a strategy for scaling a training pipeline across multiple nodes, including handling data parallelism, model parallelism, and checkpointing strategies to ensure reproducibility and fault tolerance.
LLMOps and Production Pipelines
You must architect end-to-end systems including RAG and agentic workflows. Focus on the integration of vector databases and orchestration tools. Exercise: Build a prototype RAG pipeline using a vector database and a framework like LangChain or LlamaIndex, then define a monitoring strategy using tools like MLflow or Weights & Biases to track retrieval accuracy and model drift.

Your preparation checklist

Prepare a research portfolio summary
Compile a list of your top-tier publications (NeurIPS, ICML, ACL, CVPR) and patents. For each, prepare a concise explanation of the problem, your novel algorithmic contribution, and the resulting impact or benchmark performance.
Develop an on-premise AGI research proposal
Draft a high-level roadmap for achieving on-premise agentic systems with emotional empathy and multimodal memory. Be prepared to discuss how you would balance these frontier research goals with the practical constraints of production-ready deployment.
Architect an end-to-end GenAI system walkthrough
Prepare a technical deep-dive into a past project involving RAG, prompt orchestration, or agentic workflows. Focus on your design choices for distributed training, inference optimization, and how you integrated these into a scalable cloud-based production environment.
Synthesize MLOps and LLMOps expertise
Review your experience with tools like MLflow, Weights & Biases, and cloud platforms (AWS/GCP/Azure). Prepare to discuss how you have managed data pipelines and model lifecycle management to ensure reproducibility and reliability in large-scale AI deployments.
Formulate strategic questions for the AI TIC
Prepare thoughtful questions regarding the AI Technology Innovation Centre's current research priorities, the balance between long-term scientific discovery and short-term commercialization, and the specific infrastructure available for large-scale model training.

Availability has not yet been verified. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy

KEEP EXPLORING

Similar AI jobs

Related skills and specializations in India. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…