ProductAI SolutionsLean + Kaizen Open to product roles

I design AI solutions.

AI, as in Aman Intelligence.

Product owner, not just a builder. I start with the business problem, find where AI actually earns its place, and ship solutions that make people's work measurably more efficient.

Get in touch Resume

Scroll to think out loud

How I think

Coding is not the flex. Understanding the problem is.

Anyone can be handed a spec. The value is upstream: seeing what is actually slow or wasteful in how work gets done, and designing the leanest solution that fixes it, with AI only where it earns its place.

01 Understand

02 Design

03 Ship

04 Measure

repeat · kaizen

Problem before solution

I do not start with a model or a framework. I start with what is slow, costly, or broken in how the work actually gets done, then work backwards to the smallest thing that fixes it.

AI where it earns its place

AI is a tool, not the goal. I add it where it removes real friction for a real user, and leave it out where a simpler mechanism does the job better.

Kaizen: small wins that compound

Efficiency is a habit, not a launch. Ship, watch how people actually adopt it, cut the next bit of waste. Lean thinking applied to product, not just factories.

End-to-end ownership

Research, design, build, launch, and measure. I can write the code myself, so nothing is lost in translation, but the code is the easy part. The work is understanding the problem.

7 AI & product solutions shipped

3,000+ Daily operators whose adoption I analyzed

30% Runtime cut from a core algorithm

10,000+ Users of software I have shipped

Selected work

Problems, and what I designed to solve them

Each of these started as a business problem, not a brief to build something. The tech is downstream.

2026 SampoornaCRM

AI Voice CRMReal EstateNext.js

Problem

Real-estate agents live on the phone and refuse to do data entry, so leads rot in notebooks and follow-ups quietly slip.

Solution

A voice-first CRM: the agent speaks one note in their own language and AI logs the call, updates the lead, and books the follow-up. Click-to-call, WhatsApp, listings and automation sit underneath.

Outcome

The data-entry step disappears. The CRM fills itself from how agents already work, instead of asking them to change.

Voice noteany language

AI parserapi / ai / chat

Lead updateSupabase

Follow-upcalendar + booking

NotifyWhatsApp + tasks

Agent speaks one voice note. The CRM fills itself.

2026 Trinetra

Decision SystemsDeterministic AITypeScript

Problem

When the Strait of Hormuz closed, India took 6 days to reroute crude. Decisions that big need AI speed, but a language model on its own is neither trustworthy nor auditable.

Solution

Trinetra replays the crisis and re-plans in about 4 minutes: the LLM only proposes options, a zero-LLM critic and deterministic arbiter decide, and every decision lands in a hash-chained, replayable audit trail. Runs fully offline.

Outcome

6 days of desk work compressed to minutes, with audit-grade trust: byte-identical replays and full provenance on every number.

Signal feedevent replay

Re-scoreexposure graph

LLM proposeroption cards

Critic + arbiterzero-LLM · deterministic

Audit trailhash-chained

The model proposes. Deterministic logic decides. Every step is auditable.

2026 CaseCompass

AI SimulatorEdTechPersonalization

Problem

MBA case prep is one-size-fits-all, so students burn hours drilling what they already know and skip the gaps that actually cost them offers.

Solution

An AI interviewer runs realistic simulations; engagement telemetry and percentile clustering find each student’s real weak spots and reshape the plan around them.

Outcome

Prep time gets spent where it moves the needle, personalized at scale instead of by a coach’s guess.

Studentcase journeys

AI interviewersimulation

Telemetryengagement

Clusteringpercentile ranks

Personalized plantargets real gaps

Every practice session sharpens the next one.

2026 Navkar Real Estate Academy

LMSThree PortalsZero-framework

Problem

A RERA-first academy needed to run students, teachers, and admins, but had no engineering team to maintain a heavy platform.

Solution

A deliberately lean LMS: marketing site plus student, teacher, and admin portals, with all copy and course data in plain editable files. No framework, no build step.

Outcome

A full academy platform that a non-technical team can actually run and edit day to day.

2026 Navkar Crest

AstroGSAPReal Estate

Problem

A township developer needed a launch presence for Anandam World City that felt as premium as a 180-acre integrated project.

Solution

A fast static Astro site with a considered GSAP + Lenis motion system, structured so the client can confirm copy without touching code.

Outcome

A credible digital front door for the project, quick to load and easy to keep current.

2026 Jain Earthmovers

ReactWebGL 3DIndustrial

Problem

A heavy-equipment business needed enquiries, and a generic template would have said nothing about what they actually do.

Solution

A dark-industrial React site with a real WebGL excavator in the hero that articulates and digs as you scroll, and a single content file so every spec and number stays editable.

Outcome

A site that both converts and demonstrates capability, without locking the client out of their own content.

2025 Haptic App family

iOSSwiftUIWeb API

Problem

Video is flat: it carries picture and sound but none of the physical feedback that makes a moment land.

Solution

An iOS app that records and shares video with synchronized haptics, a player for immersive playback, and a web API that brings haptics-with-sound to Android browsers.

Outcome

A small ecosystem exploring a genuinely new sensory layer for everyday video.

Experience

Where I learned to own outcomes

Apr 2026 — Jun 2026

Digital Technology Program Intern

GE Vernova · DTDP

Owned adoption analytics for a Manufacturing Execution System used daily by 3,000+ shop-floor operators across 8+ global sites.
Ran New Relic user-journey analysis to find exactly where operators dropped off, then shipped a Power BI dashboard that turned that into decisions for site leadership.
Built an AI agent that scans dashboard coverage gaps and writes site-specific insights automatically, replacing hours of manual analysis per site with automated reporting.

Aug 2024 — Jun 2025

Associate Software Engineer

MathWorks India · Data Acquisition

Cut a core data-acquisition algorithm’s runtime by 30% through profiling and redesign, directly reducing friction for enterprise clients.
Delivered a remote hardware-demonstration platform so sales could put real lab hardware in front of 25+ enterprise clients without shipping a single unit.
Shipped C++/MATLAB fixes across 3 consecutive releases used by 10,000+ customers. Promoted ahead of schedule.

Jul 2023 — Jul 2024

Associate, Engineering Development Group

MathWorks India · 50% customer-facing / 50% development

Resolved 100+ MATLAB and Simulink cases at a 90%+ resolution rate across 30+ industry domains, learning how customers actually use the tools.
Synthesized 70+ customer conversations into structured market-requirement signals that fed product roadmaps, the voice-of-customer half of product work.
Redesigned cross-team feedback workflows and cut cycle time by 40% across a 10-person group.

Education

Where the foundation was built

2025 — 2027 (expected)

MBA · Management

Department of Management Studies, IIT Delhi

Campus Finalist, American Express Campus Challenge 2025 (Decision Science): the sole team selected from DMS IIT Delhi against 500+ B-school teams.
99.74 percentile in CAT 2023 (DILR).

2019 — 2023

B.Tech · Electrical Engineering

Maulana Azad National Institute of Technology, Bhopal (NIT Bhopal)

CGPA 8.21 / 10. IEEE-published primary author: a battery-life forecasting ML model reaching 90% accuracy from State-of-Health data.
99.74 percentile in JEE Mains 2019 (Physics).

How I execute

Three languages, one owner

Most handoffs lose something between the business, the product, and the build. I speak all three, so the intent survives all the way to the shipped thing.

01 · Business

Frame the problem, size the opportunity, define what "better" means in numbers before anything gets built.

Problem framingVoice of customerRoadmap prioritizationAdoption & KPI analysisLean / Kaizen

02 · Product

Turn messy operations into flows, specs, and personas an engineer can build and a user will actually adopt.

User researchPRDs & specsWorkflow designTelemetry (New Relic, Power BI)Prompt design

03 · Build

Ship it myself when that is fastest, so nothing is lost between the idea and the working product.

Next.js / TypeScriptSupabase / SQLPythonClaude · OpenRouter · AWSAgentic workflows