AI/ML Lead

Wipro•Austin, TX

3d•Onsite

About The Position

Wipro Limited is seeking an AI/ML Lead with expertise in designing and implementing context-engineered prompts and skills on AWS Bedrock. This role focuses on transforming deterministic findings into high-quality, defensible dispositions for migration and remediation. The ideal candidate will own prompt and reasoning-skill design, build validation logic to ensure accuracy and traceability, and collaborate with Graph/RAG Engineers. The primary focus is on reasoning quality and defensibility, not the underlying infrastructure.

Requirements

Amazon Bedrock (mandatory — Claude/Titan models, Bedrock Agents, Bedrock Knowledge Bases)
Prompt engineering and context engineering
LLM reasoning and agentic reasoning design (LangGraph, CrewAI, or equivalent)
Output validation and evaluation frameworks (LangSmith, TruLens, RAGAS or similar)
Working with deterministic + probabilistic hybrid systems
Large Language Model LLMOps
6+ years in AI/ML engineering with 2+ years focused on LLM prompt/context engineering on AWS Bedrock specifically.
Experience converting rule-based or graph-based findings into LLM-assisted recommendations.
Familiarity with hallucination mitigation, confidence scoring, and human-in-the-loop review design.

Nice To Haves

Financial services domain exposure is a plus but not mandatory for this role.

Responsibilities

Design context-engineered prompts and reusable "skills" on Bedrock for dispositioning, impact analysis, and remediation recommendations.
Build evaluation and validation frameworks (e.g., LangSmith, TruLens, RAGAS-style approaches) that check LLM outputs against deterministic findings for accuracy and defensibility.
Convert deterministic (rule-based/graph) findings into natural-language, evidence-backed dispositions engineers can act on.
Identify and reduce hallucination, low-confidence, or unsupported recommendations.
Support ID de-collision reasoning — distinguishing true ID collisions from co-mingled data at the table level.
Iterate prompt/skill design across scan rounds as new patterns and edge cases emerge.