Principal Analytics Engineer · Mimecast

Anand Surendran Nair

I build LLM systems that make changes in Salesforce and other business tools.

Tiered LLM automation on AWS Bedrock and Snowflake Cortex, with confidence-based routing: the model handles the clear cases and a person reviews the rest. Built on 10+ years in analytics and production pipelines in dbt, Snowflake, and Python.

Email me LinkedIn Résumé (PDF)
Portrait of Anand Surendran Nair

Selected work

80%
of fixes needed no human review, across a 100,000-account backfill

Salesforce data cleanup agent

Finds and fixes bad account, opportunity, and quote data in Salesforce. Deterministic rules handle the clear problems. The rest go through tiered LLM routing on AWS Bedrock: Nova Lite classifies, and Nova Pro handles ambiguous cases. High-confidence fixes are written to Salesforce automatically. Low-confidence ones go to a rep, and the rep's reasoning feeds a learning run that refines the rules. Runs on 500 accounts a day.

  • AWS Bedrock
  • n8n
  • Salesforce
60%
of suggested fixes accepted without changes

Defect triage for a Salesforce migration

For a Salesforce Revenue Cloud migration. Retrieval over Jira requirements, past defects, and workaround docs (Snowflake Cortex Search, refreshed by n8n), combined with live Salesforce data, recommends a fix with a confidence level. A built-in requirements check flagged the 5% of tickets that contradicted the spec, so nobody built them as bug fixes.

  • Snowflake Cortex Search
  • n8n
  • Jira
~95%
less manual review by customer success managers

Account health summaries

Pipelines that send 6+ unstructured sources (Gong calls, Salesforce cases, NPS surveys, Nektar, Gainsight) through Snowflake Cortex LLM functions and write an automated health summary for each account.

  • Snowflake Cortex LLM
  • dbt
  • Gong
22,000+
accounts covered

From model predictions to sales tasks

Our propensity models predict which accounts are likely to expand and which are likely to buy for the first time. This pipeline turns those predictions into a self-advancing 4-stage Salesforce task chain, with checkpoint detection across multiple sources, priority locking, and incremental merge logic.

  • dbt
  • Snowflake
  • Boomi

Current role

Feb 2023 – Present
Remote

Mimecast North America Inc.

Principal Analytics Engineer, AI & Business Transformation · Jul 2026 – Present
Senior Analytics Engineer, AI & Business Transformation · Feb 2023 – Jul 2026

Other work at Mimecast:

  • Engineered 100+ ML features parsed from free-text LLM output (AWS Glue) and joined them into the feature store for three scoring models: churn, expansion, and new business. This contributed to a 10% accuracy lift.
  • Built a prospect intelligence dashboard (Python, Streamlit, Snowflake) that combines 40+ data fields with an LLM-based news service. It cut BDR research time by ~75%.
  • Built CLI tools, scheduled scripts, and self-serve Streamlit apps with Claude Code, GitHub Copilot, and the Anthropic API. I ship dbt models and Python pipelines at roughly 2x individual throughput.
  • Built an initiative-level attribution model across 45,000+ accounts, both model-predicted and organic. It stitches task progression, lead funnel stages, opportunity lifecycle, and ARR, with attribution by propensity model and rep hierarchy rollups.
Show 4 moreShow less
  • Designed the modular dbt layer (staging, intermediate, mart). It feeds Tableau dashboards, ML models, and operational analytics for 200+ users in CS, Sales, and RevOps.
  • Built dbt models that detect DMARC and Incydr license overages from product telemetry. Daily incremental loads trigger Slack and Salesforce notifications, which replaced a manual workflow.
  • Set up the data quality framework: row-count-drop tests, source freshness checks, Metaplane anomaly monitoring, and CI/CD through GitHub PR reviews. Designed the data standards from the ground up with Data Engineering.
  • Cut Snowflake compute cost through duplicate job elimination, LLM token filtering, and scheduling tuning. Built an AI adoption tracking pipeline on Anthropic's Compliance API (stored procedure on a cron schedule) across 280 licenses.

Skills

Languages & Tools

  • SQL
  • Python
  • dbt (certified)
  • Snowflake Cortex
  • AWS Bedrock
  • AWS Glue
  • BigQuery
  • Spark / PySpark
  • n8n
  • Git / GitHub
  • CI/CD
  • Docker
  • Terraform

AI & Techniques

  • AI agent design & orchestration
  • RAG
  • GenAI / LLM pipelines
  • Tiered confidence routing
  • Human-in-the-loop review
  • Kimball star schema
  • Medallion architecture
  • ELT/ETL pipeline design
  • Data quality & testing
  • Data & AI governance
  • ML feature engineering
  • KPI standardization
  • Cloud cost optimization

Platforms & Integrations

  • Salesforce
  • Gong
  • Kafka
  • Kestra
  • GCS
  • JIRA
  • Confluence

BI & Visualization

  • Tableau
  • Streamlit
  • ThoughtSpot

Education & Certifications

Education

M.S. Business AnalyticsUniversity of Cincinnati, Lindner College of Business · 2021 – 2022

B.Tech.National Institute of Technology

Certifications

dbt Developer Certificationdbt Labs

Certificate Programme in Business AnalyticsIndian School of Business