Saikiran Andey Revenue Intelligence at LatentView Analytics

I sit with client teams, build the AI SI agents and data tools they ask for, and check that the numbers hold.

Open to Forward Deployed Engineer roles. I built the MCP servers behind a Claude agent for 700 Veeam sales team members.

Ask my agent

  • 1M+

    records in the entity resolution I built

    Veeam, via LatentView

  • 3

    MCP servers with 15+ tools behind the Claude agent

    Veeam, via LatentView

  • 11

    read-only MCP tools I published on npm

    npm: tableau-lineage-mcp
  • 2,563

    automated checks in HEOR Studio, none failing

    Run the checks yourself

Projects

Four products I designed, built and shipped in 2026, and one client agent.

The lineage graph tableau-lineage.com draws for its sample workbook: parameters, calculated fields, raw fields and worksheets as linked nodes.

Tableau Workbook Auditor

Audits a Tableau workbook in your browser. The file never leaves your machine.

Ships the same engine as an 11-tool MCP server on npm.

Read the case: Tableau Workbook Auditor
HEOR Studio after loading its demo study: the generated code pane, with tabs for the SAS and the SQL it wrote and the study’s files listed in order.

HEOR Studio

An LLM drafts a reviewable study spec, and deterministic emitters write the SAS and SQL.

Verified by 2,563 checks across 24 groups, 0 failing.

Read the case: HEOR Studio
The sample case console from eigen-rho.com: a dark SQL editor with the question for one step of a case called The Vanishing SKU and the query that answers it.

Eigen-Rho

A live SQL interview platform that grades each answer by running it on Postgres.

Load-tested at 20 concurrent distinct cases: 0 errors, 14 ms p95.

Read the case: Eigen-Rho
The ClipShuttle panel’s download settings: a quality picker, the folder it saves to, the box that imports the clip into the open project, and the Download button.

ClipShuttle

An open-source Adobe Premiere Pro panel that brings YouTube footage into the open project.

Runs in Premiere 2020 to 2025 on macOS.

Read the case: ClipShuttle

Claude agent in Microsoft Teams

client workVeeam Software, through LatentView

I built and tested the MCP servers behind it.

  1. Microsoft TeamsA rep asks
  2. Claude agentMaps each rep to their accounts by Teams sign-in
  3. 3 MCP serverschurn riskcustomer 360 profileCRM records15+ tools over Databricks tables
  4. AnswerBack in Teams

Every request runs the same harness: an action, mapped to a tool, then the output.

Read it in Experience

Credentials

Awards

  • Apr to Jun 2026

    Encore Award

    LatentView Analytics

    Certificate of appreciation for outstanding performance, April to June 2026.

  • Dec 2021

    Impact Award

    Mu Sigma

    One of the first members of the “Creating Order Out Of Chaos” (COOOC) initiative; with the team, created and implemented its governance framework.

    Impact Award

    The certificate reads: “As one of the first team members of COOOC, he has collaborated with his team members to create and implement the framework for Governance.”

    The expansion of COOOC comes from my own records; the certificate uses the acronym.

  • Jan 2021

    Spot Award

    Mu Sigma

    Key contributor to the team’s 2020 automation goals.

    Spot Award

    The certificate reads: “Sai Kiran has been a key contributor to our 2020 automation goals. He has played an effective role in understanding problem statements and creating solutions for the same.”

Education

  • 2023 to 2024

    Northeastern University

    Master of Science in Business Analytics

  • 2020 to 2021

    IIIT Bangalore

    Post Graduate Diploma in Machine Learning and Artificial Intelligence

    IIIT Bangalore

    Taken alongside my years at Mu Sigma. The modules below are printed on the certificate.

    • Statistics Essentials
    • ML-1
    • ML-2
    • NLP
    • Deep Learning and Neural Networks
    • Reinforcement Learning
    • Deployment and Capstone
  • 2014 to 2018

    GITAM University

    Bachelor of Technology, Electronics and Communication Engineering

Publication

  • 2023

    Mapping Radioligand Therapy Referrals in Prostate Cancer: Leveraging Multiple Data Assets to Map the Path to Administration

    Co-author, PMSA Annual Conference 2023.

Experience

  1. LatentView Analytics

    Revenue Intelligence, client-embedded (Veeam Software, Adobe, Intuit)

    to Present

    • Translate ambiguous business problems into analytic solution requirements inside each client team; at Veeam, partner with stakeholders across 6+ sales and business segments and present findings to senior leadership.
    1. Veeam Software

      to Present

      • Found in an earlier UAT round that contracts expiring on the 29th usually roll into the following days, and added grace-period windows to the churn logic so those accounts were not flagged as churn risk by mistake.
      • Built entity resolution in SQL and Python, applying probabilistic record linkage and statistical matching across 1M+ records to improve matching accuracy by about 30% and produce the de-duplicated master dataset the revenue reporting runs on.

      Contracts expiring on the 29th were flagged as churn risk by mistake; a grace window now waits for the renewal.

      29th

      Read as a lapseGrace window

      Renewal

      Flagged as churn riskNot flagged

      4 moreShow fewer
      • Built and tested 3 MCP servers (churn risk, customer 360 profile, CRM records from Salesforce-sourced tables in Databricks) with 15+ tools behind a Claude agent in Microsoft Teams that maps each rep to their own accounts through their Teams sign-in; passed final UAT with zero errors ahead of an October 2026 rollout to 700 sales team members.
      • Migrated the revenue reporting data architecture from legacy to next-generation Salesforce, re-engineering the KPI logic and reconciling each KPI against its legacy definition while rebuilding 15+ Tableau dashboards the global sales organization uses daily, with no disruption to business-critical KPIs through the cutover.
      • Built the Databricks pipelines (ingestion, transformation, delivery) that feed the revenue dashboards and forecasting models, processing 2M+ records at 99%+ refresh reliability.
      • Embedded revenue forecasting and anomaly detection, plus natural-language querying and automated insight summaries with fallback handling for low-confidence outputs, in the reporting; together they cut stakeholder analysis and meeting-prep time by about 10 hours per week.

      In my words

      Before the agent, reps looked up churn risk in a Tableau dashboard. The agent turns that lookup into a question in Microsoft Teams. I built the MCP servers behind it, reading tables in Databricks, and a rep’s Teams sign-in maps them to their own accounts, so the agent answers only about accounts they own. Every request runs the same harness: an action, mapped to a tool, then the output. I built the servers, tested them, and got them delivered.

      When reporting moved to next-generation Salesforce, I rebuilt the dashboards one at a time and reconciled every KPI against its legacy definition through the cutover. The whole global sales division reads these dashboards daily for its updates, so the bar was that nobody lost a number during the move.

      A dashboard is only as trustworthy as the pipeline under it. A failed refresh puts a wrong number in front of an executive.

    2. Adobe

      to

      • Wrote automation scripts, using agentic AI tools, that pull CPI (cost per install) for 4 regions (AMER, EMEA, APAC, Japan) for week-over-week comparisons, and ran root-cause deep dives tying CPI swings to paid media spend for the mobile-analytics team.
      2 moreShow fewer
      • Built Power BI dashboards for MAU and CPI analysis with the mobile-analytics and lifecycle-marketing teams.
      • Built an early dormancy detection model that Lifecycle Marketing used to prioritize retention interventions.

      In my words

      I worked inside Adobe’s mobile-analytics team on paid-media measurement. Every week my scripts pulled cost per install for each of the 4 regions, compared it with the week before, and when it moved I dug into why, so marketing could see how its paid media spend was being used.

    3. Intuit

      to

      • Automated sales and inventory analytics pipelines in Python with model retraining feeding the reporting dashboards, and applied fuzzy-logic deduplication to improve data matching; ran 10+ knowledge-transfer sessions with the offshore team, backed by process documentation.
  2. Master’s at Northeastern

    to

  3. Chryselys

    Senior Analyst, Business Analytics

    to

    • Designed wireframes, then built a validation interface that automated descriptive statistics and data-quality checks on incoming client data, replacing manual review.
    2 moreShow fewer
    • Built provider identity resolution to NPI records for Prostate Radioligand Therapy providers and organizations, closing IQVIA data gaps with web scraping and fuzzy-logic matching and raising NPI mapping accuracy by 60%.
    • Built a patient-journey process tracing 10 key encounters around each therapy administration date across integrated client and IQVIA data, and co-authored the patient-journey paper presented at the PMSA Annual Conference 2023.

    In my words

    For the patient journey I traced the 10 key encounters around each therapy administration date, before and after it, across the client’s data and IQVIA’s. I also presented weekly readouts on patient trends, therapy adoption and provider networks.

  4. ZS Associates

    Decision Analytics Associate Consultant

    to

    • Took Salesforce-integrated workflows through requirements, user-acceptance testing and troubleshooting with onshore stakeholders, coordinating 5 associates.
    2 moreShow fewer
    • Set up Salesforce data-governance processes after a gap analysis with business stakeholders, with Python/VBA automation that eliminated manual data-entry errors and saved about 12 hours per associate weekly.
    • Streamlined commercial data aggregation for Key Opinion Leader (KOL) targeting, contributing to a 15% increase in targeted prescriptions in Q2 2022.
  5. Mu Sigma

    Decision Scientist

    to

    • Built data-governance and QA workflows in SAS and Python that cut errors and analysis time by over 90%, with reusable scripts adopted across multiple teams.
    2 moreShow fewer
    • Automated the recurring patient-pool extraction, cost, incidence, prevalence and survival-modeling workflows in Python and SAS, saving 1+ workday per project.
    • Built Kaplan-Meier and Cox proportional-hazards models on Flatiron oncology data (cohorts from Dx/Rx codes) comparing on-label and off-label outcomes, and helped reproduce a randomized melanoma trial’s results from real-world data.

    In my words

    If real-world methods can’t reproduce what a randomized trial found, nobody should trust them with questions a trial can’t answer. That was the point of the melanoma project. I also ran hypothesis tests comparing treatment cohorts, researched and presented real-world-evidence strategies for FDA label-extension protocols, and stratified my survival models by clinical and demographic variables.

Skills

Every skill here has a source. Open one to see where I used it.

AI Agents and MCP

  • Claude API

    • Veeam Softwareclient work

      Claude is the model behind the Teams agent I built the MCP servers for.

    • HEOR Studiochecked at build

      Its extract_spec tool calls the Claude API, and only when a key is set. Live app, opens in a new tab

    • Eigen-Rhomy word

      The next version, not live yet, uses the Anthropic SDK.

    • CCA-Fchecked live

      Claude Certified Architect, Foundations (CCA-F). Valid until Nov 11, 2026. Skilljar, opens in a new tab

  • Claude Code

    • Tableau Workbook Auditorchecked at build

      Built with Claude Code, and it ships as a Claude Code plugin. GitHub, opens in a new tab

    • HEOR Studiomy word

      Built with Claude Code.

    • ClipShuttlechecked at build

      Built with Claude Code; the README says so. GitHub, opens in a new tab

    • Eigen-Rhomy word

      Built with Claude Code.

  • MCP server design and publishing (TypeScript, Zod input schemas)

    • Veeam Softwareclient work

      3 MCP servers with 15+ tools behind a Claude agent in Microsoft Teams.

    • Tableau Workbook Auditorchecked live

      An 11-tool MCP server on npm and the official MCP Registry, written in TypeScript with Zod input schemas. npm, opens in a new tab

    • HEOR Studiochecked at build

      heor-studio-mcp on npm, with Zod input schemas. npm, opens in a new tab

  • tool API design

    • Tableau Workbook Auditorchecked at build

      Each of its 11 read-only tools reads from a local file path, so nothing is uploaded. GitHub, opens in a new tab

    • Veeam Softwareclient work

      Every request runs the same harness: an action, mapped to a tool, then the output.

    • HEOR Studiochecked at build

      Deterministic tools by default; the one model-backed tool appears only when a key is present. Live app, opens in a new tab

  • RAG

    • LatentView Analyticsmy word

      Automated minutes of meeting from week-over-week client call summaries with a RAG pipeline.

  • LLM output validation (human sign-off gates, low-confidence fallbacks)

    • HEOR Studiochecked at build

      generate_code won’t run without an explicit sign-off and every code list marked verified. Live app, opens in a new tab

    • Veeam Softwareclient work

      Natural-language querying in the revenue reporting falls back when the model’s confidence is low.

    • HEOR Studiochecked at build

      Analyses it can’t compute honestly are refused before any code is generated, with the reason. Live app, opens in a new tab

Languages

  • Python (pandas, scikit-learn)

    • Veeam Softwareclient work

      Entity resolution in SQL and Python across 1M+ records.

    • Intuitclient work

      Sales and inventory analytics pipelines in Python, with model retraining.

    • Mu Sigmamy word

      Patient-pool, incidence, prevalence and survival workflows automated in Python and SAS.

    • ZS Associatesmy word

      Python and VBA automation behind the Salesforce data-governance processes.

  • SQL (CTEs, window functions, query tuning)

    • Veeam Softwareclient work

      Entity resolution in SQL: probabilistic record linkage and statistical matching.

    • HEOR Studiochecked at build

      Emits SQL that runs in Postgres against answers worked out by hand. Live app, opens in a new tab

    • Eigen-Rhomy word

      Grades each SQL answer by running it on Postgres.

  • TypeScript

  • Node.js

    • ClipShuttlechecked at build

      A Node engine inside Premiere drives yt-dlp, ffmpeg and deno. GitHub, opens in a new tab

    • Tableau Workbook Auditorchecked at build

      The MCP server runs on Node over stdio. GitHub, opens in a new tab

    • Eigen-Rhomy word

      An Express server on Node.

  • React

  • SAS

    • Mu Sigmamy word

      Data-governance and QA workflows in SAS and Python.

    • HEOR Studiochecked at build

      Emits SAS programs as twins of the SQL. The SAS is not run on this site. Live app, opens in a new tab

  • R

    • my word

      Listed on my word. Nothing public on this page shows R.

Data Modeling and Pipelines

  • entity and identity resolution (probabilistic record linkage, fuzzy matching)

    • Veeam Softwareclient work

      Entity resolution across 1M+ records, improving matching accuracy by about 30%.

    • Chryselysmy word

      Provider identity resolution to NPI records with fuzzy-logic matching, raising mapping accuracy by 60%.

    • Intuitclient work

      Fuzzy-logic deduplication in the sales and inventory pipelines.

  • multi-tenant schema isolation (schema-per-case Postgres with read-only roles)

    • Eigen-Rhomy word

      Every case is its own Postgres schema with its own read-only role.

  • KPI and metric logic

    • Veeam Softwareclient work

      Re-engineered the KPI logic and reconciled each KPI against its legacy definition.

  • Databricks pipelines (ingestion, transformation, delivery)

    • Veeam Softwareclient work

      Pipelines processing 2M+ records at 99%+ refresh reliability.

    • Veeam Softwareclient work

      The agent’s MCP servers read tables in Databricks.

  • Salesforce data-architecture migration

    • Veeam Softwareclient work

      Moved the revenue reporting from legacy to next-generation Salesforce.

  • data governance

    • Mu Sigmamy word

      Data-governance and QA workflows that cut errors and analysis time by over 90%.

    • Mu Sigmamy word

      Impact Award: with the team, created and implemented the governance framework for COOOC.

    • ZS Associatesmy word

      Salesforce data-governance processes after a gap analysis with business stakeholders.

Solution Delivery

  • client-embedded discovery and solution design

    • LatentView Analyticsclient work

      Embedded in three client teams in a row, turning ambiguous business problems into solution requirements.

  • AI-assisted prototyping

    • Adobeclient work

      Wrote the CPI automation scripts using agentic AI tools.

    • Tableau Workbook Auditorchecked at build

      Prototyped and shipped with Claude Code. GitHub, opens in a new tab

  • UAT

    • Veeam Softwareclient work

      Final UAT with zero errors, after an earlier round surfaced contracts expiring on the 29th.

    • ZS Associatesmy word

      Took Salesforce-integrated workflows through user-acceptance testing.

  • platform cutover (Salesforce migration)

    • Veeam Softwareclient work

      No disruption to business-critical KPIs through the cutover.

  • troubleshooting

    • ClipShuttlechecked at build

      The README records the bugs that cost real debugging time, and how each was fixed. GitHub, opens in a new tab

    • ZS Associatesmy word

      Troubleshot Salesforce-integrated workflows with onshore stakeholders.

  • knowledge-transfer sessions

    • Intuitclient work

      Ran 10+ knowledge-transfer sessions with the offshore team, backed by process documentation.

  • privacy-preserving design

    • Tableau Workbook Auditorchecked live

      The workbook never leaves your machine: no backend, and the audit on this page makes no network requests.

  • Cloudflare Workers

  • Docker

    • Eigen-Rhomy word

      Ships as a multi-stage Docker image.

  • GitHub Actions CI

  • npm and MCP Registry publishing

Analytics and ML

  • Tableau

    • Veeam Softwareclient work

      Rebuilt 15+ Tableau dashboards through the Salesforce cutover.

    • Tableau Workbook Auditorchecked live

      Built an auditor for Tableau workbooks; it runs on this page.

  • Power BI

    • Adobeclient work

      Power BI dashboards for MAU and CPI analysis.

  • revenue and paid-media analytics

    • Veeam Softwareclient work

      Revenue dashboards and the pipelines under them for the global sales organization.

    • Adobeclient work

      Cost per install for 4 regions, week over week, tied to paid media spend.

  • forecasting

    • Veeam Softwareclient work

      Revenue forecasting embedded in the reporting, fed by the Databricks pipelines.

  • anomaly detection

    • Veeam Softwareclient work

      Anomaly detection embedded in the revenue reporting.

  • root-cause analysis

    • Adobeclient work

      Root-cause deep dives tying CPI swings to paid media spend.

Also on my résumé

Zod input schemas, human sign-off gates, low-confidence fallbacks, pandas, scikit-learn, CTEs, window functions, identity resolution, probabilistic record linkage, fuzzy matching, multi-tenant schema isolation, data-architecture migration, client-embedded discovery and solution design.

Building agents for customers? Let’s talk.

saiki.andey@gmail.com