AI Engineer (Agentic / Gen AI)

Larsen & Toubro Limited; Chennai; Vor Ort; INR 1.800.000 - 3.600.000; Vollzeit

Zusammenfassung

Larsen & Toubro Limited in Chennai seeks an Applied Research Engineer, AI (EPC & Manufacturing) to lead applied research on small language models and agentic AI for edge deployments. You will craft domain-specific models, build knowledge graphs, and advance multi-agent workflows across design, procurement, construction, and operations.

Responsibilities include designing, training, and deploying SLMs with low-latency edge inference, integrating digital threads, and collaborating on MLOps with

Qualifikationen

• Master's or PhD in AI/ML, CS, or related field (or equivalent experience).
• 4+ years in applied AI with hands-on experience building/training SLMs or derived models (e.g., LoRA/PEFT fine-tuning, model distillation).
• Experience deploying edge AI (e.g., NVIDIA Jetson, ARM inference, quantization techniques like AWQ/GPTQ).
• Strong experience with agentic AI (e.g., LangChain, AutoGen) and Neo4j for complex knowledge graphs (Cypher queries, graph embeddings, graph algorithms).
• Proficiency in PyTorch, Hugging Face Transformers, Neo4j, and edge frameworks (e.g., TensorRT, OpenVINO).

Aufgaben

• Design, train, and distill SLMs or derived models for domain-specific tasks in EPC/manufacturing, with deeper context engineering and mature agentic AI workflows, optimizing for low-latency inference on edge devices for site level intelligence and predictability.
• Architect agentic AI frameworks that leverage SLMs for autonomous decision-making, such as multi-agent orchestration for construction sequencing or factory floor optimization.
• Build and integrate digital threads—end-to-end data pipelines linking design, procurement, construction, and operations—using SLMs for natural language querying and automation.
• Develop complex knowledge graphs to model intricate relationships in EPC ecosystems (e.g., asset hierarchies, supplier networks, regulatory compliance), enhanced by SLM-driven reasoning and graph RAG.
• Deploy edge AI solutions for real-time applications, including IoT sensor fusion, on-device anomalydetection, and federated learning in harsh manufacturing environments.
• Prototype and benchmark systems for scalability, collaborating on MLOps pipelines with tools like Kubernetes, TensorRT, and ONNX for edge-cloud hybrid deployments.
• Conduct research on SLM fine-tuning, quantization, and agentic behaviors, documenting findings for internal adoption or publications.

Kenntnisse

Edge AI deployment

Agentic AI

Knowledge graphs

EPC/manufacturing domain knowledge

Ausbildung

Master's degree in AI/ML or related field

PhD in AI/ML or related field

Tools

LangChain

AutoGen

Neo4j

TensorRT

ONNX

PyTorch

Hugging Face Transformers

TensorRT/OpenVINO

Kubernetes

Jobbeschreibung

## AI Engineer (Agentic / Gen AI)LNT/AE-G/1896643* DC-Divisional Corporate* Chennai Head Qrs.* Posted On15 Oct 2026* End Date13 Apr 2027* Required Experience10 - 20 years## Skills#### Knowledge & Posting LocationArtificial Intelligence#### Minimum QualificationBachelor of Engineering (BE)## Job DescriptionJob Title: Applied Research Engineer, AI (EPC & Manufacturing)Focus areas : Edge AI, Agentic AILocation: Chennai, Tamil NaduJob Summary:You'll lead applied research in creating and optimizing small language models (SLMs) and derived modelsfor range of use cases cutting across both structured and unstructured data – for edge AI deployment,powering agentic systems, digital threads, and highly contextual knowledge graphs. Focus on real-worldEPC/manufacturing scenarios like autonomous workflow orchestration, connected operations, and aimedat taking digital transformation in these sectors to the next generation of technologies and withbreakthrough impact.Key Responsibilities:• Design, train, and distill SLMs or derived models (e.g., from Llama/Mistral) for domain-specifictasks in EPC/manufacturing, with deeper context engineering and mature agentic AI workflows,optimizing for low-latency inference on edge devices for site level intelligence and predictability.• Architect agentic AI frameworks that leverage SLMs for autonomous decision-making, such asmulti-agent orchestration for construction sequencing or factory floor optimization.• Build and integrate digital threads—end-to-end data pipelines linking design, procurement,construction, and operations—using SLMs for natural language querying and automation.• Develop complex knowledge graphs to model intricate relationships in EPC ecosystems (e.g., assethierarchies, supplier networks, regulatory compliance), enhanced by SLM-driven reasoning andgraph RAG.• Deploy edge AI solutions for real-time applications, including IoT sensor fusion, on-device anomalydetection, and federated learning in harsh manufacturing environments.• Prototype and benchmark systems for scalability, collaborating on MLOps pipelines with tools likeKubernetes, TensorRT, and ONNX for edge-cloud hybrid deployments.• Conduct research on SLM fine-tuning, quantization, and agentic behaviors, documenting findings forinternal adoption or publications.Sensitivity: LNT Construction Internal UseRequired Qualifications:• Master's or PhD in AI/ML, Computer Science, or related field (or equivalent experience).• 4+ years in applied AI, with deep hands-on experience building/training SLMs or derived models(e.g., LoRA/PEFT fine-tuning, model distillation).• Proven expertise in edge AI deployment (e.g., NVIDIA Jetson, ARM-based inference, quantizationtechniques like AWQ/GPTQ).• Strong experience with agentic AI (e.g., LangChain, AutoGen) and Neo4j for complex knowledgegraphs (Cypher queries, graph embeddings, graph algorithms).• Familiarity with digital threads in manufacturing/EPC: PLM integration, BIM data flows, or IIoTprotocols.• Proficiency in PyTorch, Hugging Face Transformers, Neo4j, and edge frameworks (e.g., TensorFlowLite, OpenVINO).Preferred Skills:• Experience in EPC/manufacturing domains: procurement-to-pay automation, digital twins, orsupply chain graph analytics.• Knowledge of federated learning, secure multi-party computation, or quantum-safe edge crypto.• Contributions to open-source SLM/edge AI projects or relevant publications (e.g., ICML, NeurIPSworkshops).Sensitivity: LNT Construction Internal Use