Senior Product Manager · Fintech, Payments & Applied AI

Subhasish
Goswami

I build lending and AI products: scaling international student lending, rebuilding credit decisioning around configurable rules, and shipping production GenAI. Based in Bengaluru.

subhasishgoswami02@gmail.comlinkedin.com/in/subhasishgoswami
PROFILE / 00Portrait of Subhasish Goswami

Bengaluru, India
14 years across product, lending, and analytics

01~$550M

cumulative disbursed

02~10x

origination growth, 2021 to 2025

03<10 sec

decisioning, down from 48 hrs

0498%

AI support assistant containment rate

01 / CREDIT DECISIONING

From a 48-hour queue to a sub-10-second decision

Context / decision

Loan decisions ran through a manual underwriting queue. I led the move to an automated decisioning platform built on configurable primitives: eligibility rules, pricing tiers, partner routing, and rollback, each implemented through configuration.

Verified outcome

Result: decisions in under 10 seconds, 3x throughput with no added headcount. I rebuilt the core mechanics as a public demo: Decision Studio ↗.

Scope: product owner for the decisioning platform, requirements and config logic; execution by MPOWER's in-house engineering team.

  1. Application
  2. Eligibility rules
  3. Scorecard
  4. Pricing tier
  5. Decision
    <10 sec
Every step, eligibility rules, scorecard weights, and pricing tiers, is config-driven and editable without a code deploy.
02 / AI PRODUCT

An AI assistant that handled support end to end

Context / decision

I owned product strategy and requirements for a RAG-based AI borrower-support assistant: grounding and guardrail requirements, the knowledge base, evaluation metrics for containment and hallucination, and pre-launch red teaming of grounding failures.

Verified outcome

Result: 35,277 users served through September 2025, 98% contained without a human handoff.

Scope: product strategy, grounding and guardrail requirements, knowledge base ownership, and evaluation metrics; engineering build owned by MPOWER's AI engineering team.

  1. Borrower question
  2. Intent + FAQ match
  3. RAG knowledge lookup
  4. Grounded response
98% of conversations are contained at this path with no human handoff; the remainder escalate to a live agent.
03 / CREDIT DATA STRATEGY

Cut bureau spend while extending coverage to 190+ countries

Context / decision

Analyzed $1.41M in annual credit bureau spend across 151K pulls, found 40% avoidable, and put a sign-off process in place. In parallel, began extending credit data coverage from 18 to 190+ countries through customer-supplied credit data, now in implementation with a 9-country pilot.

Verified outcome

Result: $275K per year in credit pull cost reduction, roughly $350K per year in total data and vendor cost reductions.

Scope: led the bureau spend analysis and sign-off process; partnered with risk, compliance, and data engineering on the coverage expansion.

04 / CONVERSION

Rebuilding the application funnel

Context / decision

Redesigned the loan application flow around progressive disclosure and smarter defaults, with instrumentation at every step.

Verified outcome

Result: drop-off cut from 16% to 4%, completion up from 70% to 82%.

Scope: owned the funnel redesign end to end, from instrumentation and experiment design to the shipped flow.

Drop-off16% → 4%Completion70% → 82%
Light bars: before. Blue bars: after. All figures are share of applicants at that funnel stage.
POLICY-AS-CODE PATTERN / 02

Regulated credit decision sandbox

Policy-as-code for a synthetic lending decision: adverse-action reason codes, a reproducible audit trail, and a human review path. All synthetic.

View on GitHub ↗
TOOL-USE PATTERN / 03

RAG evaluation harness

A test harness for grounded AI answers: containment, resolution accuracy, and hallucination checks.

View on GitHub ↗
REFLECTION PATTERN / 04

Multi-agent document reviewer

Line-by-line PRD and spec review from multiple stakeholder personas, built on Claude.

Private build — ask for a walkthrough

Based in Bengaluru.

14 years across product, lending, and analytics. Joined MPOWER as employee #20 in India, was promoted four times to Senior Manager, Product, and served as acting product head for both lending lines from April 2026 (MPOWER Financing, December 2019 to September 2026). Earlier TietoEVRY, Cognizant, and Wipro. MBA from Great Lakes Institute of Management; AI Product Leadership Certification, Product Faculty, September 2026; AI Product Management Certification, Product Faculty, November 2025. Evenings usually find me building with AI: the tools on this page started as weekend prototypes. I like products where a config change beats a code deploy, and metrics you can defend in a reference check.

“The first principle is that you must not fool yourself, and you are the easiest person to fool.”— Richard Feynman

I've had the absolute pleasure of working alongside Subhasish for nearly 7 years, watching him excel and grow into an incredible Senior Product Manager. We've tackled numerous technology and product rollouts together, and through it all, he has consistently been the absolute go-to person in the company for bridging the gap between engineering, business, and product teams. Subhasish is incredibly level-headed and has an unmatched knack for cross-functional collaboration and stakeholder management. When it comes to building out requirements, he always asks the exact right, detailed questions. He takes true product ownership, he knows the product inside and out, and is constantly finding ways to improve it. Whenever we hit a roadblock, his calm, structured problem-solving approach makes troubleshooting a breeze. He is fantastic with follow-up and keeps everyone aligned, always keeping his eye on driving business outcomes and pushing our company goals forward. Subhasish is a massive asset to any team he's on, and I really can't recommend him highly enough!

FBFelicia BudimanStrategic Operations & Program Delivery, MPOWER FinancingWorked with Subhasish on different teams · July 16, 2026

I have worked with Subhasish from May 2019 to December 2019 on one of our priority projects and found him to be extremely reliable. Subhasish always ensured the engineering team had the requirements they needed and followed up our business stakeholders when required, very articulate with his requirements gathering, always stepped up to take ownership of tasks and followed through to the end. It has been an absolute please working with Subhasish and would highly recommend him for product roles.

KMKylie McKiernanProduct Management, ValuePRO Software (PropTech)Worked with Subhasish at a different company · December 12, 2019
FINAL NODE / CONTACT

Building a lending or AI product?

Happy to talk.

Email me