Apramay Gyan — Product Manager

Career

Work experience

Jan 2025 – Present

Jinee Inc

Philadelphia, PA

Product Manager

Grew monthly active users 45% by defining product instrumentation, SQL-based cohort models, and a personalization framework to surface high-intent users.
Improved onboarding conversion 30% by partnering with ML engineering to translate ranking logic into actionable user stories across a three-sided platform.
Shipped a GenAI-powered self-service solution co-architected from feature scoping to production — cutting manual processing 90% and lifting client autonomy 60%.
Led multi-variable A/B experimentation across billing, risk, and compliance, lifting 30-day engagement to 60%.

Jan – May 2024

Astral Insights

Philadelphia, PA

Consultant (Capstone)

Improved profitability forecast accuracy 20% by surveying 100+ maritime professionals and building KPI frameworks to model key revenue levers.
Identified 75 strategic market opportunities via competitive landscape analysis to refine Total Addressable Market.

Jul – Dec 2023

Park Genius

Philadelphia, PA

Product Manager (Intern)

Shaped core product roadmap through 50+ user interviews, directly driving GPS tracking and real-time availability features.
Validated MVP with entrepreneurs-in-residence via Figma prototype and API documentation, securing sign-off for launch planning.

Nov 2021 – Jul 2022

Infosys Limited

Bengaluru, India

Associate Consultant

Reduced system incidents 30% and delivered $2M in cost savings migrating 200 servers via Azure DevOps & CI/CD with zero post-migration defects.
Accelerated SLA adherence 40% through cross-functional collaboration for Reckitt's SAP cloud migration.

Jul 2019 – Nov 2021

TCS

Kolkata, India

System Engineer

Maintained 99.99% uptime across 160+ countries for KPMG managing identity and access management across hybrid cloud and on-prem environments.
Improved account-creation cycle time 80% by redesigning the end-to-end workflow with a 10-member global team.

Portfolio

Case studies & projects

Data Visualization · Public Portfolio

Tableau Analytics Dashboards

Interactive dashboards spanning product analytics, market analysis, and data storytelling — each reflecting the analytical rigor I bring to product decisions.

Open dashboards

GenAI · Platform

GenAI Self-Service Platform at Jinee

Manual processing was the core bottleneck on Jinee's three-sided compliance platform — clients had to rely on internal teams for every workflow step, creating delays and scaling problems. I identified this as both a retention risk and a product opportunity, then co-authored the full technical architecture with ML engineering to replace human-in-the-loop steps with a GenAI assistant.

90%

Processing time cut

60%

Client autonomy lift

0→1

Shipped to production

My role

Product lead — feature scoping, architecture review, user story definition, launch

Problem

Clients needed human support for every action — couldn't scale, hurt retention

Solution

GenAI-powered self-service layer embedded in the platform's core workflow

Stack involved

LLMs, ML engineering, compliance workflows, billing and risk integrations

Key outcome

Clients could resolve issues independently without support tickets — dramatically reducing operational load and increasing platform stickiness.

Growth · ML

Personalization Framework — 45% MAU Growth

The platform lacked a systematic way to identify and re-engage high-intent users. I built the product instrumentation layer from scratch, then designed SQL-based cohort models that segmented users by behavior signals — feeding directly into a personalization framework that surfaced the right content and prompts at the right moment.

45%

MAU growth

30%

Onboarding conversion

My role

Defined instrumentation, built cohort logic in SQL, led ML collaboration for ranking

Problem

No visibility into user intent — couldn't differentiate active from at-risk users

Solution

Cohort models + personalized surfaces showing high-intent users the most relevant actions

Tools used

SQL, Snowflake, ML feature engineering, product analytics

Key outcome

Monthly active users grew 45% and onboarding conversion improved 30% — driven by surfacing the right moments to the right users, not just more notifications.

Experimentation

Multi-Variable A/B Framework across Billing, Risk & Compliance

No formal experimentation culture existed. I designed and led a multi-variable A/B framework that ran tests across three high-stakes workflows simultaneously — billing, risk, and compliance — without them interfering with each other. The goal wasn't just to optimize individual steps, but to identify leading indicators that predicted long-term retention.

60%

30-day engagement

Workflows covered

My role

Designed framework, defined metrics, ran experiments, analyzed results, socialized findings

Problem

Decisions made on intuition — no data to validate what drove real long-term retention

Solution

Structured A/B framework with guardrail metrics and a leading indicator model

Tools used

SQL, product analytics, statistical significance testing, Snowflake

Key outcome

30-day engagement reached 60% — but more importantly, the team now had a repeatable framework and shared language for making product bets with confidence.

Personal Project

IPO Pulse — Automated Daily Market Intelligence Tool

Built a fully automated IPO monitoring tool that fetches live market data from the Finnhub API every morning, filters for high-value listings above $200M in offer size, and delivers a formatted HTML digest straight to my inbox — no manual work, no noise. Scheduled via GitHub Actions cron job so it runs daily without any intervention. The whole thing is tested, documented, and production-ready.

$200M+

Deal size filter

Daily

Fully automated

Manual steps

What it does

Pulls live IPO calendar data, filters by deal size, and emails a formatted daily briefing automatically

How it runs

GitHub Actions cron job — triggers daily, no infrastructure to manage, zero ongoing effort

Stack

Python, Finnhub API, GitHub Actions, SMTP, dotenv — fully tested with automated test suite

Why I built it

Tracking high-value IPOs manually is noisy and slow — I wanted a zero-effort signal that only surfaces what actually matters

What it shows

This is how I think about data products — identify a real information need, eliminate the manual steps, and deliver a clean signal automatically. The same instinct drives how I build product instrumentation and alerting systems at work.

Personal Project

OrbitLog — Astronaut & Mission Tracking API

Built a production-ready REST API service for tracking astronaut profiles and spaceflight missions using an AI-driven "spec coding" approach with Cursor. I wrote the requirements doc and OpenAPI contract first, then let the AI agent work through implementation — ending up with a fully functional CRUD backend backed by Apache Cassandra via DataStax Astra DB. The whole point was to test how far precise specs could take an agentic coding workflow.

REST endpoints shipped

NoSQL tables designed

Spec-first, agent-built

What I built

FastAPI backend with full CRUD for astronauts & missions, connected to Apache Cassandra (Astra DB)

Methodology

Spec Coding — requirements → design doc → OpenAPI spec → AI-generated implementation

Stack

Python, FastAPI, Apache Cassandra, DataStax Astra DB, Pydantic, Cursor AI

Why I built it

To sharpen how I write specs and understand what agentic coding can actually deliver when given precise product requirements

Key takeaway

The quality of the output was directly tied to the quality of the spec. Writing tighter requirements and clearer API contracts made the AI agent dramatically more effective — which maps exactly to how I think about PM work.

Personal Project

Pulse — AI Product Analytics Copilot

PMs drown in raw metric dumps and spend too much time manually synthesizing data into something stakeholders can act on. I built Pulse — a working AI-powered tool that takes raw product data (funnels, A/B results, churn reports, user feedback) and instantly surfaces risks, opportunities, and next actions in PM-grade format. Built entirely end-to-end as a live demo.

Live

Working demo

Analysis modes

Claude-powered

Problem

PMs spend hours manually structuring raw metrics into insight summaries

Solution

Paste any product data → AI returns structured risks, opportunities, and recommendations

My role

Solo — ideated, scoped, designed UI, prompted the AI layer, shipped to demo

What it shows

GenAI product thinking, self-service tooling instinct, end-to-end ownership

Why it matters

This is the same GenAI self-service pattern I shipped at Jinee — applied to a problem I personally felt as a PM. Built in days, demoed in interviews, fully functional.

Turning complex
systems into
products people love.

Work experience

Case studies & projects

Skills & tools

Open to new opportunities

Turning complexsystems intoproducts people love.

Work experience

Case studies & projects

Skills & tools

Open to new opportunities

Turning complex
systems into
products people love.