Lead AI Engineer

Role: Lead Agentic AI Engineer

Experience: 5–10 years

Employment Type: Full-time

ROLE SUMMARY

We are looking for a Lead Agentic AI Engineer who can own the end-to-end design and delivery of complex, production-grade agentic systems. You will be the go-to technical expert and the engine room of our most demanding AI initiatives — turning ambiguous client challenges into scalable, functional platforms. You will drive technical solutioning for client engagements, architect multi-agent pipelines, and bridge AI engineering with business outcomes while elevating the capability of the team around you.

WHAT YOU'LL DO

Agentic Architecture & Engineering

• System Design: Architect multi-agent systems — orchestrator/sub-agent patterns, state machines, tool registries — using Microsoft Agent Framework, LangGraph, CrewAI, AutoGen, or Semantic Kernel

• Advanced RAG: Design and optimize retrieval pipelines: hybrid search, re-ranking, query expansion, multi-hop reasoning, and knowledge graphs

• Model Adaptation: Apply Quantization, PEFT/LoRA fine-tuning, and prompt optimization techniques to adapt foundation models for client-specific tasks

• Guardrails & Hallucination Control: Design and enforce comprehensive guardrail frameworks — output validation, factual grounding checks, prompt injection defenses, content filtering, and hallucination-mitigation strategies (chain-of-verification, retrieval grounding, self-consistency) — for enterprise-grade deployments

MLOps & Production Readiness

• Deployment: Productionize AI services on AWS / Azure using Docker, Kubernetes, and CI/CD pipelines (GitHub Actions / Azure DevOps)

• Observability: Build comprehensive monitoring for LLM systems — tracking accuracy, hallucinations, latency, cost, and drift using LangSmith or Arize Phoenix

• Evaluation: Define and implement LLM evaluation suites using RAGAS, G-Eval, TruLens, or custom metrics aligned to client KPIs

• Cost & Token Optimization: Drive down inference costs through token budgeting, prompt compression, KV-cache management, model routing, streaming strategies, and intelligent batching — balancing performance against cost at scale

• CI/CD: Own and evolve CI/CD pipelines for ML systems, enforcing automated testing (unit, contract, and model-quality tests) as a standard across all engagements

• Performance Tuning: Optimize model serving for high-throughput production using vLLM, DeepSpeed, or Triton Inference Server

Client Solutioning & Leadership

• Solutioning: Lead technical discovery and proposal for AI engagements; translate ambiguous client problems into actionable AI solutions

• Mentorship: Guide junior engineers, review architecture decisions, and build the team’s internal library of reusable AI patterns, accelerators, and playbooks

• Stakeholder Communication: Present solution designs, demo prototypes, and communicate technical trade-offs clearly to client technical and business stakeholders

MUST-HAVE QUALIFICATIONS

• Experience: 5–10 years in software engineering or data science, with at least 3 years in applied Gen AI / LLM engineering in a services or consulting context

• Agentic Frameworks: Proven experience building production agents with LangGraph, CrewAI, AutoGen, or Semantic Kernel

• RAG & Retrieval: Deep expertise in RAG architectures, vector databases (Pinecone, Qdrant, Weaviate), and embedding pipelines

• LLMs: Strong working knowledge of GPT-4o, Claude 3.x/4.x, Gemini, and open-source models (Llama 3, Mistral)

• Cloud & DevOps: Hands-on with AWS / Azure AI services; Docker, Kubernetes, and CI/CD workflows

• Engineering: Strong Python, FastAPI, SQL; software design patterns; a “software engineering first” approach to ML — with rigorous unit, integration, and model-quality testing

• MLOps & Productionization: Proven track record taking LLM systems from prototype to production — owning deployment pipelines, observability, evaluation suites, guardrails, and ongoing model health in live client environments

• Education: B.Tech / B.E. / M.Tech in Computer Science or related discipline

GOOD TO HAVE

• Fine-tuning: Experience with PEFT/LoRA fine-tuning workflows and serving optimized models

• Performance Tuning: Hands-on experience with vLLM, DeepSpeed, or Triton Inference Server for high-throughput model serving

• Knowledge Graphs: Exposure to GraphRAG or ontology-based retrieval strategies

• Multi-modal: Experience with vision-language models or multi-modal agent pipelines

• Certifications: AWS Solutions Architect, Azure AI Engineer Associate, or equivalent