Sr_ Development Lead

Job description

Minimum 5+ Years in Software / Machine Learning (with 1.5+ years hands-on in LLMs & GenAI) Core Tech Stack: Python, PyTorch, LangChain, LlamaIndex, Vector DBs, FastAPI, Docker, Cloud AI (AWS/Azure) Target Work Model: Hybrid (Enterprise Office / Remote flexibility available) Employment Type: Full-Time, Permanent Position We are seeking a talented, technically passionate Generative AI Engineer with 3+ years of professional software engineering and applied AI/ML experience to design, build, and deploy production-grade LLM applications. In this hands-on engineering role, you will turn the latest advancements in generative artificial intelligence into resilient, scalable business solutions. You will specialize in constructing Advanced Retrieval-Augmented Generation (RAG) pipelines, engineering autonomous agents with deterministic tool-calling capabilities, optimizing prompt workflows, experimenting with parameter-efficient model adaptation (LoRA/QLoRA), and serving high-performance inference APIs via containerized microservices hosted on AWS or Microsoft Azure. Key Responsibilities RAG Architecture & Vector Search Engineering RAG Pipeline Development: Design and implement end-to-end RAG pipelines incorporating multi-modal document parsing, recursive text splitting, and semantic chunking. Vector Store Management: Integrate, optimize, and benchmark vector databases (Pinecone, Qdrant, Chroma, Weaviate, pgvector, Azure AI Search) for sub-second similarity search. Hybrid Search & Re-ranking: Implement hybrid retrieval combining dense vector embeddings with sparse lexical keyword matching (BM25) and cross-encoder re-ranking models (Cohere Rerank, BGE). Context Optimization: Engineer contextual compression, parent-document retrieval, and metadata filtering schemas to eliminate context noise and hallucinations. B. Agentic Workflows & Multi-Tool Integration Agentic Frameworks: Build stateful multi-agent workflows and autonomous task execution loops using LangChain, LangGraph, LlamaIndex, or AutoGen. Tool Calling & Action APIs: Develop robust, type-safe function-calling and tool-use interfaces connecting LLMs to external REST APIs, relational databases, and enterprise systems. Loop Safeguards & Routing: Implement plan-and-solve agent patterns, multi-step reflection loops, and fallback error-handling to prevent infinite agent recursion. C. Prompt Engineering, Output Validation & Fine-Tuning Prompt Engineering: Design and test systematic prompt templates utilizing Chain-of-Thought (CoT), Few-Shot prompting, ReAct patterns, and directional stimulus prompting. Structured Output Parsing: Enforce deterministic, validated output structures using Pydantic, Instructor, Guidance, or Outlines to guarantee reliable downstream ingestion. Model Adaptation: Conduct Parameter-Efficient Fine-Tuning (PEFT, LoRA, QLoRA) on open-weights models (Llama 3, Mistral, Phi-3/4) for domain-specific tasks. D. Backend API & Inference Microservices Inference APIs (FastAPI): Develop high-throughput asynchronous backend microservices using Python (FastAPI), managing connection pooling, concurrency, and rate limiting. Streaming Latency: Implement low-latency streaming inference pipelines utilizing Server-Sent Events (SSE) and WebSockets for conversational user interfaces. Inference Engines: Optimize inference throughput utilizing modern serving engines like vLLM, TensorRT-LLM, or Ollama, minimizing Time-to-First-Token (TTFT). Containerized Deployment: Containerize applications with Docker and configure deployment manifests for Kubernetes, AWS ECS/EKS, or Azure Container Apps. E. Evaluation, Guardrails & LLMOps Observability Automated Model Evaluation: Establish automated evaluation benchmarks utilizing synthetic test datasets and evaluation harnesses (Ragas, TruLens, DeepEval) to quantify faithfulness and recall. AI Guardrails & Safety: Implement enterprise guardrails (NeMo Guardrails, Llama Guard, Azure Content Safety) to proactively mitigate prompt injection, jailbreaks, and toxicity. Observability & Tracing: Integrate comprehensive LLM tracing, prompt versioning, and cost monitoring using tools like Langfuse, Phoenix, or Weights & Biases. Skill Requirements Required Technical Skills & Qualifications Total Experience: Minimum 3+ years of professional software engineering / data science experience, with at least 1.5+ years of dedicated hands-on GenAI / LLM development. Education: Bachelor’s degree in Computer Science, Software Engineering, Artificial Intelligence, Data Science, or equivalent technical field. Python & ML Foundations: Advanced Python programming skills (asyncio, type hints, Pydantic, pytest) and deep familiarity with PyTorch or Hugging Face Transformers. GenAI Frameworks: Extensive hands-on experience with LangChain, LangGraph, LlamaIndex, and commercial/open foundation model APIs (OpenAI, Anthropic, Bedrock). Vector Search & RAG: Proven track record building and deploying production RAG architectures using vector databases (Pinecone, Chroma, Qdrant, pgvector). Backend & Containers: Hands-on proficiency developing asynchronous REST APIs using FastAPI, Docker containerization, and Git-based collaborative workflows. Core Competencies: Strong analytical problem-solving skills, solid understanding of transformer architecture mechanics, and proactive technical curiosity. 4. Preferred Qualifications & Industry Certifications Inference Optimization: Hands-on experience with model quantization (GGUF, AWQ, GPTQ) and high-throughput inference serving engines (vLLM, TGI, TensorRT-LLM). Cloud Certifications: Cloud AI certifications: AWS Certified AI Practitioner / ML Specialty, Microsoft Certified: Azure AI Engineer Associate (AI-102), or Azure AI Fundamentals. Open Source & Portfolio: Demonstrated contributions to open-source GenAI projects, LangChain/LlamaIndex community plugins, or active technical blog/GitHub repository. Agile Collaboration: Experience working in Agile/Scrum sprint environments collaborating with Product Managers, UX Designers, and Frontend Developers. Other Requirements . Candidate Evaluation Matrix (For Technical Interview Panels) Technical interviewers should evaluate candidates across the following five core technical engineering competencies: Evaluation Area Competency Focus Key Technical Signals RAG Architecture & Retrieval Chunking strategies, vector search, hybrid retrieval, and re-ranking Explains concrete trade-offs between chunk sizes; demonstrates practical experience with hybrid search (BM25 + vector) and cross-encoders to improve recall. Agentic Loops & Tool Calling Function calling, stateful graphs, and error recovery mechanisms Can code and debug an agent with deterministic tool use; understands state management in LangGraph; handles schema parsing errors gracefully. Prompt Engineering & Parsing Few-shot prompting, structured JSON schema validation (Pydantic) Demonstrates advanced prompting techniques (ReAct, CoT); knows how to force LLMs to emit valid JSON schemas reliably using Instructor or Outlines. Backend & Microservice APIs FastAPI, asynchronous streaming, connection pooling, and Docker Comfortably writes async Python endpoints streaming tokens via SSE; explains how to containerize an application and manage GPU/CPU resource constraints. Evaluation & LLMOps Guardrails Synthetic benchmarking, hallucination metrics, and guardrail layers Articulates how to evaluate an LLM app using Ragas or TruLens; explains how NeMo Guardrails or regex/semantic filters prevent prompt injections. Apply now Information at a Glance Why HCLTech? At HCLTech, you'll supercharge your potential. You'll find your career. And you'll find your spark. All at a place that knows that helping its customers stay on top starts by putting its people first. HCLTech is a global technology company, home to more than 230,000 people across 60 countries, delivering industry-leading capabilities centered around digital, engineering, cloud and AI, powered by a broad portfolio of technology serv