Dr. Santhoshkumar

Senior ML Engineer | Senior Data Scientist

I turn machine learning ideas into enterprise AI systems that customers trust and businesses rely on.

At Eurowings (Lufthansa Group), I work with stakeholders through the full cycle—from ideation to production systems that millions of users depend on daily. I balance accuracy, latency, cost, and maintainability in every AI system I build.

Highlights

  • First scalable AI application serving 4 million customers daily
  • Fully automated MLOps improving CSAT from 32% → 74%
  • Introduced AI Agents, CI/CD, and modern ML practices to the team
  • My strength: designing AI systems that balance accuracy, latency, cost, and maintainability
LLM & RAGAI AgentsSemantic SearchMLOps & LLMOpsDatabricksAzureProduction ML
ResumeBlog

Eurowings Digital GmbH • Europe | Remote

Dr. Santhoshkumar

Production AI Leadership

Senior ML Engineer / Data Scientist

Current Focus

LLM, RAG, AI Agents, MLOps & LLMOps

Production
13+
Years

End-to-end enterprise application delivery
2012-2016: Backend (Java) Engineering → Team Lead (Enterprise Engineer)
2016-2020: PhD (ML) + Team Leadership (Research & Leadership)
2020-2022: Enterprise Analytics (Hitachi) (ML & Backend)
2022-2026: Enterprise AI Systems (Lufthansa Group)

5
Flagship Systems

Search, RAG, MLOps, Forecasting, AI Agent. 91% recall@10, 40ms P99 latency.

Proven Impact

  • First scalable AI application serving 4 million customers daily
  • Fully automated MLOps improving CSAT from 32% → 74%
  • Introduced AI Agents, CI/CD, and modern ML practices to the team

Enterprise AI Systems

Production AI systems built for enterprise scale. From search to RAG to MLOps infrastructure.

⭐ Flagship Project
🔍Eurowings · Lufthansa Group • 4-month production rollout

AI-Powered Semantic Search

Enterprise search powered by embeddings, retrieval, and LLMs

💡 Why it matters

Keyword-only search caused poor retrieval for natural-language questions, multilingual queries, and support intent inside booking journeys.

Built an onsite search and FAQ discovery platform for airline e-commerce, handling natural-language search, multilingual queries, and hybrid retrieval at customer scale.

🏗️ Architecture Highlights

  • •Nuxt.js website and backend search API
  • •Databricks query understanding for language detection, spelling correction, and intent extraction
  • •Hybrid BM25 + vector retrieval with confidence scoring and reranking
Semantic SearchRAGVector SearchGPT-4LlamaIndexDatabricks

📈 Impact

  • ✓110% search satisfaction improvement (42% → 90%)
  • ✓91% recall@10 achieved
  • ✓99.3% latency reduction (6s → 40ms P99)
  • ✓€420K annual savings (18% fewer support calls)
Read Case Study →
💬Eurowings · Lufthansa Group • 3-month production rollout

Enterprise AI Customer Support

Conversational AI for enterprise customer support

Designed and deployed a customer-support chatbot over enterprise help content, reducing repetitive questions while maintaining citation-backed responses.

🏗️ Architecture Highlights

  • •Azure OpenAI GPT-4 with retrieval-augmented answer generation
  • •Hybrid search over FAQs, policy documents, and support content
  • •Session-aware conversation memory for follow-up questions
RAGGPT-4LlamaIndexConversational AIAzure

📈 Impact

  • ✓Reduced repetitive tickets
  • ✓Faster resolution
  • ✓Higher self-service rate

💡 Core Challenge

Customers repeatedly asked about baggage, cancellations, refunds, check-in, and special assistance, overloading support channels.

Read Case Study →
🛠️Eurowings · Lufthansa Group • 6-month platform build

Enterprise MLOps Platform

Infrastructure for operating ML at scale

Built the platform layer for governed ML deployment on Azure Databricks, standardizing CI/CD, registry workflows, monitoring, and access control across teams.

🏗️ Architecture Highlights

  • •Git and Azure DevOps for versioned pipelines
  • •Databricks MLflow tracking and model registry
  • •Unity Catalog and Feature Store for governance
MLOpsDatabricksMLflowUnity CatalogAzure DevOps

📈 Impact

  • ✓Weeks → Hours deployment
  • ✓Standardized deployments
  • ✓Improved governance

💡 Core Challenge

Teams deployed models inconsistently, with weak governance, fragmented monitoring, and no reproducible delivery path.

Read Case Study →
📊Eurowings · Lufthansa Group • 5-month production rollout

AI Forecasting & Personalization

Predictive modeling for business decisions

Applied forecasting and recommendation pipelines to improve planning accuracy, targeting quality, and campaign efficiency across digital operations.

🏗️ Architecture Highlights

  • •Feature pipelines for reusable forecasting inputs
  • •Ensemble forecasting with XGBoost, Prophet, Random Forest, and regression models
  • •Collaborative and similarity-based recommendation components for personalization
XGBoostProphetCollaborative FilteringTime Series

📈 Impact

  • ✓Improved forecast accuracy
  • ✓Better targeting
  • ✓Reduced manual effort

💡 Core Challenge

Manual forecasting and non-personalized campaigns created planning inefficiencies and weaker customer targeting.

Read Case Study →
🔍Open Source • 2-month experimentation

AI Search Agent

Agentic AI combining search, reasoning, and tool calling

Built an AI agent that combines semantic search, multi-step reasoning, and external tool calling to answer complex information requests beyond traditional document retrieval.

🏗️ Architecture Highlights

  • •User query → Planner → Search Tool → Retriever → LLM Reasoning → Response Generation
  • •LangChain/LlamaIndex framework for agent orchestration
  • •Vector search for semantic retrieval
AI AgentsSemantic SearchTool CallingLangChainGPT-4

📈 Impact

  • ✓Modern agent architecture
  • ✓Reusable reference implementation
  • ✓Accelerates LLM experimentation

💡 Core Challenge

Traditional search engines return documents, but users increasingly expect conversational answers and autonomous task execution requiring multi-step reasoning.

Read Case Study →

Career Journey

13+ years building ML systems: from research to enterprise AI platforms at scale

PhD in Machine Learning

VIT, Vellore · 2020 · GPA 9/10

Completed doctoral research on rumor detection and control with deep learning.

10+ Years Team Leadership

From Software Teams to ML Teams · 2016-Present

Led software delivery teams (2016-2020), then transitioned to leading ML engineering teams building production AI systems (2020-present). Coordinated across product, engineering, and support organizations to deliver customer-facing AI platforms.

13+ Years in Industry

From Java/AEM to Production AI

Progressed from enterprise software engineering to ownership of production AI systems.

Key Leadership Responsibilities

  • Technical lead for enterprise AI initiatives involving product managers, software engineers, data scientists, and business stakeholders.
  • Designed end-to-end AI architectures from requirements gathering through production deployment and monitoring.
  • Standardized reusable MLOps patterns using Azure Databricks and MLflow to accelerate ML delivery.
  • Collaborated across engineering and business teams to translate customer problems into production AI solutions.
  • Mentored engineers on production ML practices, evaluation methodologies, and system design.
⭐

Current

Ongoing

Eurowings Digital GmbH

📍Europe
Current Role
Senior ML Engineer | Senior Data Scientist

📅July 2022 – Present

Led end-to-end development of 5 production AI systems across aviation and e-commerce domains. Architected semantic search (91% recall@10, 40ms P99), RAG solutions, MLOps/LLMOps pipelines, and AI agents. Achieved €420K annual savings through 18% support call reduction and 110% search satisfaction improvement. Technical lead for AI strategy and cross-functional delivery.

Semantic SearchRAGAI AgentsMLOps & LLMOpsAzure DatabricksTeam Leadership
2020

2020 – 2022

Hitachi Vantara (Hitachi Data Systems)

📍Pune, India
Senior Consultant (SC2)

📅July 2020 – June 2022

Contributed to enterprise data analytics initiatives. Led cross-functional teams across analytics and machine learning models. Successfully delivered enterprise analytics solutions for global customers and led multidisciplinary teams to complete projects within scope and timelines.

JavaAWSAWS EC2Scikit-learnML ModelsAEMTeam Leadership
2016

2016 – 2020

Relevance Lab

📍Bangalore, India
Team Lead

📅April 2016 – July 2020

Led a team of 10+ professionals across project planning, execution, and delivery. Coordinated stakeholder communication and mentored team members. Ensured quality standards and delivery timelines were consistently met.

JavaKafkaDockerAWSAWS S3AWS EC2Team Leadership
2014

2014 – 2016

Sapient

📍Bangalore, India
Associate Technology L2

📅September 2014 – April 2016

Developed web applications using Adobe Experience Manager (AEM). Built backend services with Java and front-end solutions with JavaScript, HTML, and CSS. Contributed to enterprise digital experience platform delivery.

JavaAWSAEMJavaScriptHTMLCSS
2012

2012 – 2014

Cognizant

📍Chennai, India
Programmer Analyst

📅December 2012 – August 2014

Developed enterprise Java applications and web services. Built scalable backend systems and contributed to full-stack development projects for global enterprise clients.

JavaAWSAEMWeb Services
13+
Years
5
Companies
5
Systems Shipped
10+
Team Leadership

🎓Education

Academic foundation in ML research and software engineering

Ph.D. in Information Technology

Vellore Institute of Technology

2014 – 2020

📍 India

Doctoral research in Machine Learning with focus on rumor control in online social networks using neural networks and bio-inspired algorithms.

Master's Degree in Software Engineering

Vellore Institute of Technology

2007 – 2012

📍 India

Advanced study in software engineering, algorithms, and system design.

Open Source & Engineering

Experimentation, tooling, and data engineering projects demonstrating breadth and curiosity.

🔧 DevTools

NOPC

Developer productivity tool for project automation

Created a reusable CLI utility to simplify repetitive project setup and automation tasks, demonstrating clean software engineering practices.

✨ Impact

  • ✓Reduced setup effort
  • ✓Cross-platform compatibility
  • ✓Extensible architecture
PythonCLIAutomationDevTools
View on GitHub →
📊 Data Engineering

Web Scraping API Experiments

Data engineering experiments for AI pipeline data collection

Comparative experiments evaluating different web scraping APIs for reliability, cost, JavaScript rendering, and anti-bot handling to identify suitable approaches for LLM and RAG data pipelines.

✨ Impact

  • ✓API benchmarking
  • ✓Cost-quality tradeoffs
  • ✓Production-ready evaluation
PythonData EngineeringWeb ScrapingAPI Integration
View on GitHub →
📊 Data Engineering

OpenAQ API to Dataset Pipeline

ETL pipeline for environmental data ingestion

Built a modular ETL pipeline to automate ingestion and preprocessing of OpenAQ environmental data, transforming raw API data into analysis-ready datasets for ML workflows.

✨ Impact

  • ✓Automated dataset creation
  • ✓Reduced preprocessing effort
  • ✓Production ETL design
PythonETLAPI IntegrationData Processing
View on GitHub →

More projects on GitHub

Tech Stack

Capabilities organized by system responsibility rather than a flat keyword list.

Generative AI & LLMs

GPT-4Azure OpenAILlamaIndexLangChainRAGPrompt EngineeringVector SearchSemantic Search

ML Engineering

PythonPyTorchTensorFlowScikit-learnXGBoostMLflowModel Serving

MLOps & Platform

DatabricksAzure MLDockerKubernetesCI/CDModel MonitoringUnity Catalog

Data & Search

Vector DatabasesAzure AI SearchElasticsearchSQLNoSQLData Pipelines

Cloud & Infrastructure

AzureAWSFastAPIREST APIsMicroservices

Latest Writing

Engineering judgement from production search, enterprise AI delivery, and retrieval system design.

2026-07-07•8 min read

Why Vector Search Never Says "No Results"

A production lesson on false positives, confidence thresholds, and why semantic retrieval needs explicit failure handling.

Vector SearchSearch EvaluationRAG
2026-07-01•9 min read

How LLMs Can Benefit from Self-Generated Context: Memento

A practical synthesis of Memento research and why context compression matters for KV cache, throughput, and cost in long-running reasoning systems.

LLMContext EngineeringInference Optimization
View all writing →

Interested in building reliable AI systems?

I'm always happy to discuss Enterprise AI, Search, RAG, MLOps, and Production Machine Learning.

Let's connect.

Send EmailLinkedIn

santhoshramuk@gmail.com • Europe | Remote