Open to product roles

I build the
systems behind
smart products.

I'm Anwesha, an AI Product Manager with full-stack engineering roots. I went from shipping C#/.NET APIs to owning product outcomes and scaling 0→1 AI systems in B2B SaaS fintech. I care about the impact of when AI automation is needed and the tools and stack to build them.

See my work ↓Let's connect

APIs deployed
in production

0 wk

Platform shipped
from 0→1

Support tickets cut
via AI notifications

A little more about me

Not your typical PM.

Pivoted from software engineering to product after a production bug had me asking a bigger question: are we even building the right thing? That curiosity landed me at a high-growth Series A in NYC, reducing the friction of onboarding agents caused by miscategorized credential data from a national verification system across all 50 states. I turned that into a user-centric workflow portal with role-based, license-aware access that helps users connect 2x faster. 0+ agents onboarded, driving a 10x increase in distribution capacity.

When I'm not building, you'll find me exploring an entire world of cuisine, getting heavily influenced by social media to try yet another matcha spot (yes, I wrote an essay about it), and 3x-ing my productivity with AI tools and automation.

Featured case study✓ Complete

Responsible AI — FairScreen

How I'd turn an academic finding about hiring bias into a shippable product spec.

Problem

Insight

What I built

Outcome

Problem

Bias hides in the aggregate.

A Stanford study (Bommasani et al., FAccT 2026) ran AI screening across 4M job applications. In aggregate, bias was invisible — it only surfaced when measured position by position.

Insight

Fairness can't be a compliance afterthought.

If you only audit at the end, you've already shipped the harm. Monitoring has to be a core product surface — live, and per-position — not a quarterly report.

What I built

A full PRD for FairScreen.

A hiring-assessment layer with live four-fifths-rule dashboards per position, aligned to Title VII and the EU AI Act's high-risk classification. Auto-pause when a ratio drifts red, audit trails by default.

Outcome

A spec that treats fairness as a feature.

Two regulatory frameworks mapped directly to product requirements, with guardrails designed in — not bolted on after launch.

View PRD on Notion →View Research →

Selected work

Things I've built.

Not mockups. Not hypotheticals. Real systems, deployed and running.

● Live

AI Automation — Marketing Analytics

End-to-end AI marketing analytics around TalentFlow, a B2B SaaS talent-assessment platform. Custom landing page with GA4 + GTM, a 21,346-row synthetic dataset, and a pipeline that turns Claude API outputs into structured Looker Studio tables.

GA4GTMLooker StudioClaude APINetlify

Landing Page →Dashboard →GitHub →

● Live

AI Automation — Payment Recovery

An AI-powered payment-recovery workflow. Stripe captures failures, n8n orchestrates webhooks, Claude API runs risk analysis and drafts recovery emails, HubSpot manages contacts, and Slack gets real-time alerts.

Stripen8nClaude APIHubSpotSlack

GitHub →

● Live

LLM Knowledge Engineering — Ledger

Inspired by Karpathy's LLM-Wiki pattern. Ledger compiles raw payments sources (Visa, Mastercard, Stripe) into a living, interlinked knowledge base — so a compliance lead or PM can trace a mandate to its downstream flows in seconds.

PythonClaude APID3.jsReactKnowledge Graphs

Live Site →Medium →GitHub →

Cinema Through the Years

A Tableau storybook analyzing domestic and international gross earnings of American films over time — global trends, director patterns, and the rise of movie popularity across countries and languages.

TableauData AnalysisData Viz

View Project →

Crime Pattern Prediction

Exploratory analysis of 2021 Denver crime statistics, with an Adaboost classifier estimating the likelihood of potential crimes from real-time locations. Focused on the impact of COVID-19 and the MeToo movement.

PythonMLAdaboost

View Project →

How I work

Build it like you own it.

I don't hand off specs and walk away. I stay close to the data, close to the tools, and close to the outcome.

Understand the Problem

Talk to stakeholders. Dig into data. Identify what's actually broken versus what people think is broken.

Map the System

Architecture first. What tools exist? What connects to what? Where does data flow and where does it break?

Build. Test. Ship.

Do the simple thing that works. Get it live, get it in front of people, get real feedback. Perfection is the enemy of shipped.

Measure It.

Dashboards that tell the truth. Insights pipelines that surface what matters. Course-correct on what the numbers say, not what feels right.

Build Guardrails.

Now that we're using AI, the game has changed. Automate intelligently, but always build the guardrails. Compliance, accuracy, and human oversight aren't optional.

Experience

Where I've worked.

Technical Business Analyst / Product Manager

Afficiency · New York City Metropolitan Area

Apr 2024 – Oct 2025

Scaled product from 100-agent pilot to 5,000+ agents across multiple locations with 40% fewer regression bugs
Discovered 38% merchant onboarding drop-off via SQL funnel analysis, shipped A/B test that improved conversion from 62% to 78%, generating $800K in additional ARR
Owned product discovery through 40+ merchant interviews, shipped AI-powered notification system that increased DAU 20% and reduced support inquiries 28%
Mentored all incoming team members as primary onboarding lead, accelerating ramp time across the organization

Product StrategySQLA/B TestingUser ResearchAI/MLMentorship

Technical Product Management Intern

Snyder Tech · Remote

Jun 2023 – Aug 2023

Drove product roadmap execution through async collaboration, writing user stories with clear acceptance criteria
Delivered 8% efficiency improvement in feature delivery while navigating evolving stakeholder priorities

Product RoadmapUser StoriesAgileStakeholder Management

Graduate Research Assistant

University of Maryland · Maryland, United States

Feb 2023 – Jun 2023

Managed a team of 8, conducted statistical analysis (regression, hypothesis testing, sampling) with 78% success rate
Oversaw project monitoring processes ensuring on-time delivery

Statistical AnalysisTeam ManagementResearchPythonExcel

Technology Analyst

Infosys · On-site

Oct 2021 – Jun 2022

Managed release lifecycle for 50+ deployments, improving system uptime from 98.2% to 99.7% and reducing error rates by 40%
Led UAT across 200+ test scenarios, reduced post-launch issues by 60% and improved release quality scores from 78% to 94%
Architected platform with 860+ RESTful APIs in C# and .NET Core, handling 2M+ transactions monthly

C#.NET CoreREST APIsRelease ManagementUATSQL

Software Engineer

Novac Tech · India

2018 – 2019

Optimized data processing bottlenecks affecting 2M+ daily records, improving system efficiency by 50%
Built automated solutions using Python and SQL, improving data accuracy from 92% to 99%

PythonSQLData ProcessingAutomation

Master of Science in Information Systems

University of Maryland, Robert H. Smith School of Business

United States

Information and Telecommunication Engineering

SRM University

India

Beyond work

The human bits.

💻

Conference Circuit

Grace Hopper Celebration in Orlando. AI Summit in NYC. Always learning, always connecting, always in the front row.

🗽

NYC Explorer

Been to every hidden and mainstream skyline view on the internet. Yes, the city has the world's best pizza. And yes, a 40-minute line for trending ice cream is absolutely a hobby.

🍵

The Matcha Essay

I fell down the matcha rabbit hole and did what any reasonable person would do. Wrote an entire Medium essay about it. Part love letter, part cultural deep dive.

Read on Medium →