Vendor Hiring Guide

ML/AI Engineer & ML Lead

Candidate Screening & Submission Guidelines

1. Purpose

This document provides detailed guidance to staffing partners and vendors for identifying suitable candidates for the ML/AI Engineer and ML Lead positions.

Over the past several interview cycles, we have observed that many candidates have resumes containing up-to-date AI technologies (e.g., Vertex AI, Gemini, ADK, LangChain, RAG), but are unable to demonstrate hands-on implementation experience during technical interviews.

The goal of this document is to clearly define the expected technical competencies and reduce profile mismatches.

2. Role Overview

We are looking for hands-on AI Engineers, not architects or delivery managers.

The ideal candidate should be capable of:

- Designing production-grade AI solutions
- Developing AI applications using Python
- Building and deploying AI Agents
- Implementing RAG pipelines
- Working extensively on Google Cloud Platform
- Mentoring junior engineers
- Working directly with US stakeholders

This is not a research role, nor a project management role.

3. Candidate Profile

Requirement

Preferred

Experience

612 years

AI/ML Experience

4+ years

GenAI Experience

2+ years

Python

Strong hands-on

GCP

Strong hands-on

Client Communication

Required

Team Leadership

Preferred for ML Lead

4. Mandatory Technical Skills

A. Python Development (Mandatory)

Candidates should be actively writing production code.

Expected proficiency includes:

- Object-Oriented Programming
- FastAPI
- REST APIs
- Exception Handling
- Logging
- Async Programming
- Collections
- Generators
- List Comprehensions
- File Processing
- Clean Code Principles

Interview Validation

Candidates will be required to complete a live Python coding exercise.

Examples include:

- Dictionary aggregation
- List processing
- Data manipulation
- Algorithmic problem solving

Candidates unable to demonstrate coding ability are unlikely to be shortlisted.

B. Google Cloud Platform (Mandatory)

Candidates must possess hands-on experience with GCP.

Expected services:

- Vertex AI
- Vertex AI Endpoints
- Vertex AI Agent Builder
- Vertex AI Search
- BigQuery
- Cloud Storage
- Cloud Run
- Cloud Functions
- IAM
- Cloud Logging
- Cloud Monitoring
- Secret Manager
- Artifact Registry

Candidates should clearly explain:

- deployment
- monitoring
- scaling
- security
- authentication

C. Gemini

Hands-on implementation experience with:





- Gemini Flash
- Gemini Pro
- Gemini Enterprise

Expected knowledge:

- Prompt Engineering
- Function Calling
- Structured Outputs
- Context Windows
- Temperature
- Top-P
- Top-K

D. Agent Development Kit (ADK)

Hands-on ADK implementation is highly preferred.

Candidates should understand:

- Agent
- Tool
- Session
- Memory
- Multi-Agent Systems
- Workflow
- Agent-to-Agent Communication
- SessionService
- Debugging
- Deployment

Simply mentioning ADK on a resume is insufficient.

E. Retrieval-Augmented Generation (RAG)

Candidates should have implemented production RAG systems.

Expected topics:

Ingestion

- PDFs
- Office Documents
- HTML
- Databases

Chunking

- Recursive
- Semantic
- Parent-child
- Chunk overlap

Embeddings

- Google
- OpenAI

HuggingFace Vector Databases

Examples:

- Vertex AI Vector Search
- Pinecone
- Qdrant
- Chroma
- Weaviate

Retrieval

- Similarity Search
- Hybrid Search
- Metadata Filtering
- Re-ranking

Prompt Construction

Response Generation

F. RAG Evaluation

Candidates should understand:

- Groundedness
- Faithfulness
- Context Precision
- Context Recall
- Answer Relevancy

Preferred tools:

- Vertex AI Evaluation
- RAGAS
- DeepEval
- LLM-as-a-Judge

G. Traditional Machine Learning

Expected knowledge:

Algorithms

- Random Forest
- XGBoost
- Logistic Regression
- Linear Regression
- Clustering

Evaluation

- Precision
- Recall
- F1
- ROC-AUC

Imbalanced Data

- SMOTE
- Class Weights
- Threshold tuning

Feature Engineering

Hyperparameter Tuning

Cross Validation

H. Production Engineering

Candidates should have experience with:

- Docker
- CI/CD
- GitHub Actions
- Cloud Build
- Deployment Pipelines
- Logging
- Monitoring
- Model Versioning
- Rollbacks
- Cost Optimization
- Latency Optimization

5. Leadership Expectations (ML Lead)

The ML Lead role is not a delivery manager position.

Expected responsibilities include:

- Architecture reviews
- Code reviews
- Technical mentoring
- Production troubleshooting
- Design discussions
- Customer interaction





Candidates should still spend a meaningful portion of their time coding.

6. Preferred (Bonus) Skills

- LangChain
- LangGraph
- MCP (Model Context Protocol)
- CrewAI
- AutoGen
- Dialogflow CX
- Prompt Optimization
- Fine-Tuning
- LoRA
- PEFT
- Document AI
- OCR
- Knowledge Graphs

7. What We Will Validate During Interviews

Every shortlisted candidate should be able to explain:

One production project end-to-end

Why architectural decisions were made

Trade-offs considered

Challenges faced

Production failures

Performance optimization

Monitoring

Deployment

8. Interview Process

Stage

Focus Area

Resume

Discussion

Production project ownership

Architecture

End-to-end AI solution design

GCP

Vertex AI, Gemini, Cloud services

ADK

Agent implementation

RAG

Architecture and optimization

ML

Algorithms and evaluation

Coding

Live Python assessment

Leadership

Mentoring and stakeholder management

9. Common Reasons for Rejection

Based on recent interview trends, the most common reasons candidates are rejected include:

- Unable to explain projects beyond a high-level overview.
- Resume contains technologies (ADK, Vertex AI, RAG, LangGraph, etc.) but lacks hands-on implementation experience.
- Weak Python coding skills or inability to complete a live coding exercise.
- Limited understanding of production deployment, monitoring, and cloud-native AI architectures.
- Strong delivery management background but limited recent hands-on technical contribution.
- Weak understanding of RAG implementation, evaluation, or traditional ML concepts. Limited or no experience with GCP, Vertex AI, Gemini, or AI agent frameworks.

10. Candidate Self-Assessment Checklist

Before submitting a profile, vendors should verify that the candidate can confidently answer "Yes" to most of the following:

Question

Yes/N o

Have you personally built and deployed a production GenAI application?

Can you explain an end-to-end RAG architecture you implemented?

Have you worked hands-on with Vertex AI and Gemini?

Have you built or deployed AI agents (preferably using ADK)?

Can you explain chunking, embeddings, vector databases, and retrieval strategies?

Have you worked with CI/CD and production deployments?

Can you complete a live Python coding assessment?

Are you currently hands-on with coding (not just reviewing code)?

Have you mentored engineers while remaining technically involved?

📌 Ai Ml Engineer (Telangana)
🏢 Quess
📍 Telangana