Boston · AI Systems · Available

Anuj Ursal

Enterprise AI, Revenue Analytics, and Products Built from Scratch

I turn messy systems into intelligent work.

I sit where finance, data, and AI meet. I built enterprise AI agents at a $1.5B+ ARR SaaS company, revenue analytics that Board decks ran on, and personal AI products shipped from scratch.

At a glance

Boston, MAEnterprise AIRevenue AnalyticsProduct BuilderKaseyaOpen to opportunities
Contact LinkedIn Kaseya case study

The focus

I work at the intersection of finance, data, and AI — translating messy operations into intelligent systems.

$1.5B+
ARR SaaS company where I built enterprise AI agents and revenue analytics
7+ yrs
across data analytics, financial systems, and AI deployment
2
personal AI products designed, built, and shipped from scratch
0
shadow AI incidents under the governance framework I owned

The path

Mumbai → New York → Boston

Each stop added a layer: data, then financial systems, then revenue at scale, then AI. The throughline is the same — complex systems, decomposed until they make sense. Click any stop for detail.

25+ KPIs dashboarded22% marketing ROI lift via ML15 hrs/wk saved with Python

First layer: raw data into decisions. Built KPI dashboards leadership actually used, applied ML to lift marketing ROI, and automated away repetitive analysis with Python.

Systems: Python · SQL · dashboarding · ML-driven marketing analytics

$1.2M annual savings, 1,800+ users98% UAT pass · 150+ cases50K transcripts → 12% churn cut

Second layer: financial systems at enterprise scale. Drove digital transformation work that saved $1.2M annually across 1,800+ users, designed UAT that passed at 98% across 150+ cases, and mined 50K call transcripts to cut churn 12%.

Systems: financial systems · UAT design · Jira traceability · transcript analytics

90% revenue variance reduction$2.5M unrecognized ARR found40% faster close

Third layer: revenue at scale. Untangled the revenue data behind a $1.5–2B ARR business — variance down 90% ($10M to under $1M), $2.5M of unrecognized ARR surfaced, month-end close 40% faster.

Systems: Power BI · SQL (Snowflake) · NetSuite · Salesforce · ARR/MRR analytics

40% ticket deflection74→94% accuracy in 30 days0 shadow AI incidents

Fourth layer: AI in production. Handed a blank canvas — implement AI enterprise-wide, no playbook, no team. Shipped a parent-child agent system in Copilot Studio, drove accuracy from 74% to 94% in 30 days with 1,000+ structured tests, and built governance people actually trusted.

Systems: Copilot Studio · Power Automate · RAG design · evaluation frameworks · AI governance

Claude architect — in progressReality Mode — liveKitchen Pro — in dev

Current layer: going deep on the model layer through the Anthropic Claude Certified Architect track while shipping two personal AI products and staying close to the Boston AI community.

Systems: Claude API · Next.js · TypeScript · Supabase · evaluation frameworks

The defining chapter

Two roles. One company.
Both moved real numbers.

Eighteen months at Kaseya — first untangling revenue data, then handed a blank canvas: implement AI enterprise-wide. No playbook, no team.

Track A — Revenue Analytics

The revenue engine

90%
monthly variance reduction — $10M to under $1M, sustained across close cycles
$2.5M
unrecognized ARR surfaced during analysis — nobody asked; it was in the data
40%
faster month-end close — 10–12 days down to 6–7
6M→559K
null subscription records remediated via mapped ID logic
Track B — Enterprise AI Agents

The agent system

40%
ticket deflection across deployed agents
74→94%
accuracy in 30 days — 1,000+ structured tests
3 agents
HR, AI Governance, IT — parent-child orchestration in Copilot Studio
Zero
shadow AI incidents — governance people actually trusted; 200+ employees trained
Read the full case study →

Products & builds

I build because I can't not

Side projects are how I keep my hands on the frontier — full products, not tutorials. Designed, built, and used by me.

craving → reflection → AI grounding → 5-min timer → outcome log
Next.jsTypeScriptOpenAI APITailwindBuilt with Codex
Claude VisionNext.jsBuilt with Claude Code

That's all I'm sharing for now.

intake → parent agent → child agent routing → governed response → eval loop
Copilot StudioPower AutomateRAGEval frameworks
Full case study →

Capabilities

The systems I build with

Six domains, one practice: AI that works, numbers that hold, products that ship.

Cluster 01

AI Agents & Automation

Agent orchestrationCopilot StudioPower AutomateClaude API · Claude CodeOpenAI APIClaude VisionRAG designPrompt engineering
Cluster 02

Evaluation & Governance

Evaluation frameworksAdversarial testing · red-teamingAccuracy & hallucination scoringAI governance policyShadow-AI controls1,000+ structured testsUAT design
Cluster 03

Revenue Analytics

Power BI · DAXSQL — Snowflake · PostgreSQL · SQL ServerPython — Pandas · NumPyTableauExcel modeling
Cluster 04

Financial Systems

NetSuiteSalesforceARR / MRR analyticsRevenue reconciliationMonth-end close
Cluster 05

Product Building

Next.jsTypeScriptThree.js · WebGLTailwindSupabaseGitVercel · Netlify · CloudflarePWA
Cluster 06

Delivery

Jira — traceability matricesConfluenceLucidchartAgile & hybridSprint planning

Currently

Staying sharp

A

Anthropic Claude Certified Architect

In progress — deep on the model layer: agent architecture, RAG, evaluation harnesses, prompt caching, and production guardrails.

B

Building products

Reality Mode live and used daily; Kitchen Pro in development. Full products, designed and shipped solo.

V

Venture Cafe @ MIT, Cambridge

Thursdays in Kendall Square — I go to connect with founders, researchers, and people solving real problems, and to keep a pulse on what's being built.

O

ODSC East 2026 · Boston

Two days deep in AI evaluation, agent reliability, and production LLM systems — sharpening how I test, measure, and ship AI that holds up in the real world.

Open to opportunities

Let's build something that matters.

Open to roles across AI enablement, finance analytics, and senior product/BA work — anywhere in the US, remote or on-site.

aursal006@gmail.com (774) 708-2013 LinkedIn Case study

resilient — by design