AGM - IT Analytics

Role Title

AGM -  AGM – AI Engineer

Position No

Function

Technology – IT

Sub Function/ Vertical/ Department

Data & Analytics

Band

M2

Reports to Role (Position No)

VP – Data & Analytics

Location

Pune

Date of last update/approval

1. Job Purpose (In one or two sentences)

• To design, build, and optimize large-scale AI solutions and data lakes by leveraging advanced Big Data frameworks (PySpark, Impala, Iceberg) to drive intelligence across enterprise operations.
• To spearhead the tuning, optimization, and prompt engineering of Small Language Models (SLMs) and Large Language Models (LLMs) to deliver high-performance, cost-effective, and highly context-aware enterprise AI workflows.

2. Key Accountabilities / Key Result Areas (Max 5)

• High-Performance AI Engineering: Build and deploy scalable AI pipelines and workflows natively utilizing PySpark, Apache Iceberg, and Impala architectures (p. 2) to process massive enterprise data streams for real-time model ingestion.
• GenAI Model Optimization: Own the end-to-end evaluation, hyperparameter tuning, and fine-tuning of SLMs and LLMs to meet internal latency, accuracy, and operational cost benchmarks.
• Prompt & Context Engineering: Architect advanced, robust, and secure prompt engineering frameworks, system instructions, and multi-turn agentic logic patterns to prevent hallucination and maximize reasoning capabilities.
• Data Lakehouse Management: Drive optimization strategies for storage layouts within Apache Iceberg, handling schema evolution, hidden partitioning, and time-travel querying to optimize deep learning training pipelines.
• Cross-Functional Governance & Industry Best Practices: Collaborate with Managed Services Partners, infrastructure leads, and external technical forums to establish stringent operational SLAs for AI application uptime, inference limits, and security protocols.

3. Core Competencies, Knowledge, Experience, Technical / Professional Qualifications (Max 5)

• 7+ Years of Experience in Domain (Telecom/IT/OEM) with 3-5 experience in handling Large Scale Telecom Network Domain
• Knowledge of service performance KPIs/SLA benchmarking and strategies for continual performance improvement.
• Domain Experience: 10+ Years of total experience in IT/Telecom/OEM domains, with a minimum of 5–7 years specifically handling Large-Scale Data Engineering and AI/ML model production ecosystems (p. 2).
• Big Data Mastery: Strong hands-on expertise with PySpark for distributed processing, Apache Iceberg for transactional lakehouses, and Impala for high-performance interactive SQL querying on massive datasets.
• Generative AI Expertise: Deep technical understanding of transformer-based architectures, model quantization techniques (GPTQ/AWQ), fine-tuning approaches (LoRA/QLoRA), and advanced context window scaling for both LLMs and resource-constrained SLMs.
• Framework Fluency: Proven track record in orchestrating production-grade workflows using frameworks like LangChain, AutoGen, LlamaIndex, or CrewAI combined with custom vector databases and Graph databases .

4. Key Performance Indicators (Max 5)

• Network Quality of Service (QoS) Adherence: Adherence to strict latency, jitter, and packet loss targets across the IP transport domain.
• Capacity & Bandwidth Utilization: Optimization of link utilization percentages to prevent backhaul congestion during peak traffic hours
• Model Inference Efficiency: Adherence to strict response latency, throughput, and token-cost targets across deployed enterprise GenAI apps.
• Data Processing & Pipeline Latency: Percentage reduction in data-ingestion cycle times through optimized PySpark execution and Iceberg partition layouts.
• Model Accuracy & Precision: Achievement of >95% accuracy/F1-scores or alignment matrix evaluation scores across customized enterprise LLM/SLM tasks.
• System Availability & Scalability: Adherence to AI system uptime SLAs, parallel execution scale limits, and Mean Time to Repair (MTTR) metrics for operational pipeline failures Query Execution Performance: Optimization of Impala analytics engine utilization percentages  during heavy concurrent retrieval requests

5. Annual Budget Owned / Key Quantitative Parameters like Workforce managed etc.

6. Risks, Challenges, Job Context (Short Description)

• Challenges:

• Managing hyper-volume, multi-structured telecommunication text, voice, and network telemetry logs under strict privacy parameters.
• Navigating fast-evolving AI trends to systematically upscale infrastructure for newer concepts like Agentic AI workflows, hybrid RAG (Retrieval-Augmented Generation), and real-time streaming pipelines

• Critical Risks

• Model Hallucination & Data Leakage: Risks associated with poor prompt alignment, leaking proprietary enterprise data via LLM prompts, or providing wrong automated resolutions to end consumers.
• Lakehouse Partition Sluggishness: Data scale outstripping query performance limitations due to suboptimal Iceberg structure layouts or inefficient Spark cluster allocations