Skip to content
Vijay Kumar
Portrait of Vijay Kumar

Product & Solutions

Vijay Kumar

I make enterprise software work for the people who have to use it.

Most enterprise software breaks at the seams between business logic and technical execution. I design and deliver the systems that close that gap.

Selected work

From promise to proof

Four agentic systems I took from a painful problem to a working, cloud-deployed vertical slice — each with full architecture, simulated enterprise backends, and a test suite that proves the claims.

Agentic AI · Service operations

Refund Assurance — the system that reopens a case when “refund processed” was a lie

When an AI service session closes with “your refund has been processed,” this agent waits out the processing window, checks Order Management, and reopens the case — only on confirmed failure.

The problem

Async refunds fail silently at the gateway; the session still counts as resolved, and nobody knows until the customer calls back.

  • TypeScript
  • React
  • Zod
  • Vitest
  • Playwright
Live demo Architecture & README

The proof

  • 73/73

    tests passing

  • 1

    bounded, idempotent, permission-checked write

  • 8

    step pipeline with independent read-after-write proof

ScopeSalesforce-style systems simulated and labeled; no model called at runtime — the LLM slot is a proven, schema-guarded boundary.

Background

Where I learned it

Client-facing fintech, then enterprise delivery, then product strategy, UX, and analytics. Open a row for detail.

  1. Work from analyst discovery through implementation to rebuild the property data behind real-estate investment decisions, and the Power BI views analysts use to make them.

    • Python
    • REST APIs
    • Snowflake
    • Power BI
    • UX research
    Related work : Rebuilding the property data analysts decide with

    Buffalo, NY · Remote

  2. Taught graduate students technical and business concepts, then turned customer interviews, JTBD research, and behavioral analytics into prioritized onboarding and UX recommendations.

    • Teaching Assistant, School of Management· Aug 2024 – Dec 2024
    • Product Strategy & UX Intern· Jan 2025 – May 2025
    • Customer interviews
    • JTBD
    • Google Analytics
    • Figma
    Related work : Turning customer research into product decisions

    Buffalo, NY

  3. Grew from turning manufacturers' quote-to-order problems into requirements and integrations to owning solution delivery end to end across Salesforce, SAP, and Configit CPQ.

    • Solutions Consultant· Feb 2021 – Feb 2023
    • Senior Solutions Consultant· Mar 2023 – May 2024
    • Salesforce
    • SAP ERP
    • CPQ / CLM
    • REST · SOAP · GraphQL
    • Azure
    • UAT
    Related work : Making quote-to-order reliable for global manufacturers

    Chennai, India

  4. Solved institutional clients' market-data and portfolio-analytics problems by working backward to the data, calculation, or platform behavior underneath, then helped cut client onboarding time by 40%.

    • Associate Solutions Consultant· Dec 2018 – Feb 2020
    • Client Solutions Consultant· Feb 2020 – Feb 2021
    • Market data
    • Portfolio analytics
    • SQL
    • Salesforce

    Hyderabad, India

Education and credentials

Education

  • University at Buffalo

    2025

    M.S., Management Information Systems (STEM)

  • MNIT Jaipur

    2018

    B.Tech., Computer Science & Engineering

My journey

FactSet taught me to work backward from what a client was seeing to what the system was actually doing. Configit put that to work in enterprise delivery for global manufacturers, and graduate study at Buffalo and fellowship work at Third Estate Ventures added the product side: research, UX, and analytics.

It reads like three careers, but it's one skill used at different layers of the same problem. I treat enterprise problems with end-to-end ownership—from the first client discovery call to the final database transaction. Whether architecting solution blueprints for complex CPQ rollouts or leading cross-functional delivery, I focus on building reliable systems that scale long after the handoff.

How I operate

  1. 01Start with the person, not the ticket

    A request is usually the visible part of the problem, so I start with the user's actual context.

    During a CPQ rollout, reps kept re-entering quotes the pricing engine rejected. Sitting with them on live quotes showed the failures came from how requirements were written, not from user error.

    EvidenceCustomer research at UB
  2. 02Trace it to the system

    I follow what someone experiences down to the data, workflow, or integration that produces it.

    When analysts reported one property appearing in several rows, I traced the records from the municipal APIs through the pipeline to where they were matched. Rebuilding that matching around how analysts actually decide cut duplicate records by 60%.

    EvidenceProperty data rebuild
  3. 03Own it past the handoff

    Requirements aren't the finish line. I stay through design, build, UAT, go-live, and production.

    I stayed on a quote-to-order go-live through UAT and the first production order cycle, watching integration logs for failures. That follow-through caught problems that would have blocked orders in the first week after launch.

    EvidenceQuote-to-order at Configit
  4. 04Let evidence set the priority

    Behavioral data, research, and measured outcomes decide what comes first, not the loudest opinion.

    Two onboarding screens looked like the main problem until analytics showed most users leaving at a later step. Ranking the recommendations by that data moved the roadmap to where people actually dropped off.

    EvidenceBehavioral analytics at UB

Let's work together.

Open to high-leverage Solution Engineering, Product Management, and Technical Architecture roles where complex systems meet commercial outcomes.