Gen AI Architect

LTM; Khordha; Hybrid; INR 1.500.000 - 2.500.000; Vollzeit

Zusammenfassung

LTM in Odisha, India is seeking a GenAI lead to architect and optimize local deployment of open-source LLMs, including distillation and multiagent orchestration. You will drive RAG pipelines, agent frameworks, and explainability standards to enable compliant, efficient AI-powered solutions.

You will shape evaluation frameworks and governance around prompts, safety, and bias while coordinating across data, engineering, and product teams in a fast-paced environment.

Qualifikationen

• Experience building GenAI agent systems and RAG solutions.
• Proficient in large language model finetuning and distillation.
• Familiar with SHAP, LIME and explainability frameworks.

Aufgaben

• Lead selection and optimization of open-source LLMs for local deployment.
• Lead finetuning using PEFT LoRA and QLoRA and full finetuning strategies.
• Architect LLM distillation pipelines for smaller domains.
• Design and govern RAG pipelines for chatbots, grievance redressal, helpdesk automation and document summarization.
• Build reusable AI agent frameworks: task agents, reasoning agents, retrieval agents, workflow managers.
• Architect multiagent orchestration systems with supervisor-worker patterns and self-reflection loops.
• Design ML pipelines for calibration, ensembles, stacking, boosting and uplift causal models.
• Establish XAI frameworks with SHAP, LIME, Integrated Gradients for interpretability.
• LLM Evaluation frameworks with ROUGE, BERTScore, LLM-as-judge.
• Implement hallucination detection and bias evaluation using LangKit, GuardrailsAI and custom evaluators.
• Define evaluation benchmarks aligned to use cases and post-finetuning evaluation.
• Govern prompt engineering standards, few-shot templates, and chain-of-thought frameworks.
• Document intelligence pipelines including OCR, parsing, embeddings, summarization, QA.

Kenntnisse

GenAI Strategy

LLM Fine-tuning

PEFT LoRA

RAG pipelines

Agent frameworks

Multiagent orchestration

Explainable AI

Model evaluation

Tools

Llama 3

Mistral

Falcon

BLOOM

LangGraph

AutoGen

PEFT

QLoRA

Jobbeschreibung

Key Responsibilities - GenAI Agentic AI Strategy

Select and optimize opensource LLMs Llama 3 Mistral Falcon BLOOM for local deployment lead finetuning using PEFTLoRAQLoRA and full finetuning strategies

Architect LLM distillation pipelines to create smaller domainspecific models for inference efficiency

Design and govern RAG pipelines standard RAG Graph RAG Hybrid RAG for chatbots grievance redressal helpdesk automation and document summarization

Build reusable AI Agent frameworks task agents reasoning agents retrieval agents workflow managers using LangGraph and AutoGen

Architect multiagent orchestration systems with supervisorworker patterns tooluse agents and selfreflection loops

Classical ML Advanced Analytics

Design ML pipelines for calibration models ensemble models stacking blending boosting upliftcausal ML models and simulation models

Establish XAI Explainable AI frameworks using SHAP LIME Integrated Gradients define interpretability standards for regulatory compliance

LLM Evaluation Quality Frameworks

Create accuracy and evaluation frameworks for summarization ROUGE BERTScore LLMasjudge extraction classification and conversational AI

Implement hallucination detection factuality scoring and bias evaluation frameworks using LangKit GuardrailsAI and custom evaluators

Define LLM evaluation benchmarks aligned to use cases establish continuous evaluation pipelines triggered post finetuning

Govern prompt engineering standards fewshot template libraries and chainofthought reasoning frameworks

Document Intelligence Knowledge Pipelines

Build document intelligence workflows OCR Tesseract PaddleOCR parsing chunking strategies embeddings summarization extraction QA