AI/ML Engineer – RAG & Retrieval Systems (Kolkata)

WBE Consultants LLC is a US-based technology and consulting firm specializing in enterprise digital transformation, with a focus on SAP S/4HANA migrations. Our India development arm, Platinum Consulting & IT Solutions Pvt Ltd, is responsible for building our flagship products.

Our product suite includes AMIGO (AI Managed Implementation Governance Office), a Salesforce-native project governance platform, and Belden, an AI-powered project intelligence agent that provides health analysis, risk intelligence, automated reporting, and decision support for complex enterprise programs.

We are looking for an AI/ML Engineer to join our team building Beldens AI engine. You will work alongside a Senior AI/ML Engineer, contributing to the development, testing, and optimization of our RAG (Retrieval-Augmented Generation) pipeline on AWS.

Belden is built entirely on AWS (Bedrock, Lambda, S3, Pinecone) and serves as the intelligence layer for AMIGOs Salesforce-based governance data. The core technical challenge is building a production-grade RAG pipeline that can accurately retrieve and reason over deeply hierarchical, relational business data.

This is an excellent opportunity for someone with foundational AI/ML experience who wants to go deep on RAG systems and work on a genuinely hard problem making retrieval work over complex enterprise data structures. Youll learn from experienced engineers while contributing meaningfully to a commercial product.

Build and maintain data transformation pipelines that convert Salesforce JSON into embedding-ready formats

Implement chunking logic that creates self-contained, contextually rich documents from hierarchical data

Develop and test Lambda functions for data ingestion, transformation, and retrieval

Maintain incremental sync processes between Salesforce (via S3) and Pinecone

Execute retrieval quality tests and document results

Build and maintain evaluation datasets (query-answer pairs with ground truth)

Implement automated testing pipelines for retrieval accuracy

Analyze retrieval failures and propose improvements to the senior engineer

Experiment with embedding models, chunking strategies, and reranking approaches

Configure and maintain Bedrock knowledge bases and agent components

Monitor Lambda performance, costs, and error rates

Implement logging and observability for pipeline debugging

Support deployment and testing across development and production environments

Develop and refine prompt templates for Beldens five core topics

Test prompt variations and document which approaches produce better outputs

Implement guardrails and scope controls to prevent out-of-domain responses

Create test suites for regression testing prompt changes

Work closely with the Salesforce development team on data format requirements

Document pipeline configurations, test results, and operational procedures

Participate in code reviews and architecture discussions

Communicate progress and blockers clearly to the team

24 years in software engineering with exposure to AI/ML, NLP, or data engineering

Hands-on experience with at least one RAG or LLM-based project (production or significant prototype)

Familiarity with the RAG pipeline concept: embedding vector store retrieval generation

Python: Strong proficiency this is your primary working language for Lambda functions and data pipelines

AWS Fundamentals: Working knowledge of S3, Lambda, IAM basics, CloudWatch logs

Vector Databases: Familiarity with Pinecone, Weaviate, or similar (experience with any vector DB is acceptable)

LLM APIs: Experience calling LLM APIs (OpenAI, Anthropic, Bedrock, or similar) and handling responses

Data Transformation: Comfortable working with JSON, handling nested structures, and writing transformation logic

Curiosity about how things work you dig into why something failed, not just that it failed

Attention to detail retrieval quality depends on careful implementation

Clear written communication youll document findings and explain technical issues to the team

Willingness to learn RAG is a fast-evolving field; you should enjoy staying current

AWS Bedrock experience: Familiarity with Bedrock agents, knowledge bases, or model invocation

Pinecone specifically: Experience with Pinecone indexing, querying, and metadata filtering

Evaluation frameworks: Experience with RAG evaluation tools (RAGAS, TruLens, or custom evaluation pipelines)

Prompt engineering: Demonstrated ability to craft prompts that produce consistent, well-structured outputs

Salesforce or CRM data: Familiarity with Salesforce object structures or similar CRM/ERP data models

LangChain or similar: Experience with LLM orchestration frameworks (helpful for understanding patterns, though we use custom code)