Product Specialist (T)

• *Forward Deployed Engineer Archetype:** Builder · Learner · Pair programmer Role Summary

You are an on\-the\-ground builder who writes real production code inside client environments from day one. You come in hungry, move fast, and turn ambiguous problems into working prototypes within hours — not weeks. Your credibility comes entirely from working software, not slides. What You Will Do

• Embed directly at client sites to prototype and deploy agentic AI workflows using frameworks such as LangGraph, CrewAI, AutoGen, or AWS Bedrock Agents — shipping working code, not slide decks.
• Build RAG pipelines end\-to\-end: chunking strategies, vector store configuration (Pinecone, pgvector, Weaviate), retrieval tuning, and response evaluation.
• Instrument LLM\-powered applications with observability tooling (LangSmith, Braintrust, Arize) so clients can see exactly what their agents are doing in production.
• Participate actively in daily client stand\-ups and technical reviews, communicating clearly about progress, blockers, and trade\-offs with both engineers and business stakeholders.
• Rapidly iterate on prototypes based on user feedback — from zero to demo in 24–48 hours is the expectation, not the exception.
• Document deployment architectures, prompt engineering decisions, and integration patterns so knowledge persists after you rotate off an engagement.
• Contribute reusable agent templates and accelerators to Cognizant's internal AI toolkit between engagements. Technical Foundation
• Strong Python; basic TypeScript / JavaScript
• REST API design and integration
• Git, CI/CD basics, containerisation (Docker)
• SQL and at least one cloud platform (AWS / Azure / GCP)
• Hands\-on LLM experience (OpenAI, Anthropic, Gemini APIs) GenAI / Agentic AI Requirements
• Has built at least one end\-to\-end RAG or agent application — personal projects count as strongly as work experience
• Understands prompt engineering, few\-shot design, and chain\-of\-thought prompting
• Familiar with agentic orchestration concepts: tool use, memory, planning loops
• Knows how to evaluate LLM output quality — even informal logging or manual review frameworks What Makes You Stand Out
• You have shipped something real with AI — a GitHub repo, a side project, a hackathon win — not just certificates
• You are comfortable being wrong in front of a client and pivoting immediately

You ask 'what does done look like?' before writing a single line of code