Shashank Nampalli

AI Infrastructure · ML Platforms · Data Platforms · Production Scale

BuildingProduction-ScaleAI and ML Systems

I'm Shashank, and I build the unglamorous-but-essential plumbing behind modern AI — the recommendation engines, the data pipelines, the GenAI infrastructure. The stuff that processes tens of terabytes a day and serves hundreds of millions of users. If that sounds like your kind of conversation, you're in the right place.

MLOpsRecommendation SystemsGenAI / LLMOpsDistributed Data
Shashank Nampalli, Senior Data and ML Engineer

30+

TB / day processed

115M+

users supported

20+

ML strategies

35%

compute reduction

Scroll

About

Engineering platforms that power AI at scale

With 10+ years of experience, I design and own production-scale systems spanning ML platform engineering, distributed Spark pipelines, recommendation infrastructure, cloud-native data platforms, and enterprise LLMOps. I focus on architecture that balances velocity, reliability, and cost at global scale.

Production-scale ML & data systems
Distributed Spark & cloud pipelines
Recommendation & vector infrastructure
GenAI / LLMOps orchestration

30+TB/day

Data processed daily

115M+

Customers supported

20+

Recommendation strategies

35%

Compute reduction

60%

Faster onboarding

Featured Projects

Production systems with measurable impact

Large-scale case studies across recommendation platforms, vector search, GenAI workflows, and enterprise data modernization.

Case Study 01

Multi-Tenant Recommendation Platform

Configuration-driven personalization at global scale

Architected a configuration-driven recommendation platform powering 20+ personalization strategies across Walmart International markets.

PySparkDataprocBigQueryAirflowCosmos DBGCS
  • 30+ TB/day processing
  • 60% reduction in onboarding effort
  • 35% compute savings

Case Study 02

Semantic Similarity & Vector Search

Embedding-based retrieval at production scale

Built embedding-based retrieval pipelines using multimodal AI models for semantic recommendation systems.

CLIPSentenceTransformersVespa DBPySpark
  • Low-latency vector retrieval
  • Improved recommendation relevance
  • Production-scale ANN search

Case Study 03

LLM-Powered Complementary Item Engine

GenAI workflows with intelligent reranking

Designed GenAI recommendation workflows using LLM-generated complementary queries and intelligent reranking systems.

GeminiPySparkAirflowBigQuery
  • Improved CTR
  • Better recommendation diversity
  • Production-grade LLMOps

Case Study 04

Enterprise Revenue Recognition Modernization

Teradata to BigQuery at 115M+ customer scale

Led migration of Verizon revenue recognition systems from Teradata to BigQuery for 115M+ customers.

BigQueryTeradataPySparkSQL
  • Reporting latency: 12h → 4h
  • 400–600M records/day
  • Modernized enterprise analytics

Case Study 05

Automated Reconciliation & Audit Platform

Enterprise-grade financial data validation

Built enterprise reconciliation systems validating consistency across multiple financial data layers.

PythonSQLBigQuery
  • 90% reduction in reconciliation defects
  • Automated auditability
  • Enterprise-grade reliability

AI & Agentic Systems

AI-native engineering workflows

From Cursor-assisted development to autonomous monitoring agents and production LLMOps — building systems where AI amplifies engineering velocity without sacrificing reliability.

AI-Assisted Engineering

Accelerating platform delivery with Cursor, Claude, and Codex workflows integrated into daily development.

agent-shell
$ |

Prompt Orchestration

Structured prompt templates, evaluation harnesses, and versioned prompt registries for production LLM features.

agent-shell
$ |

LLMOps Pipelines

End-to-end observability for embeddings, retrieval, reranking, cost tracking, and quality regression gates.

agent-shell
$ |

Autonomous Monitoring Agents

Agents that detect pipeline drift, SLA breaches, and data quality anomalies with automated remediation playbooks.

agent-shell
$ |

AI Observability

Tracing LLM calls, token economics, hallucination signals, and retrieval hit rates across recommendation surfaces.

agent-shell
$ |

Agent Orchestration

Multi-agent workflows for code review, incident triage, and data platform operations with human-in-the-loop gates.

agent-shell
$ |

orchestration.status

Multi-agent pipeline active

retrievererankevaluatedeploy

Technical Expertise

Deep stack across data, ML, and cloud

PySpark95%
Spark SQL92%
BigQuery94%
Airflow90%
SQL96%

Career Timeline

A decade of platform engineering

Work With Me

Let's Build Scalable AI Systems

Open to senior engineering roles, platform consulting, and collaborations on production AI infrastructure. Book a call or reach out directly.

Email MeDownload ResumeConnect on LinkedIn

svarma.de@gmail.com

Or send a message