Pranit Kumar
Product Manager  /  Product Thinker  /  Business Analytics
Open to PM / TPM opportunities
Pranit Kumar

Pranit
Kumar

Product Manager building AI products — my NL-to-SQL system at Fyxt hits 88% accuracy across 17,000 properties.

Product Manager  ·  Technical Program Manager  ·  Data Analyst

I bridge analytics and product strategy — from predicting NFL plays with deep learning to building AI products at Fyxt used across 17,000 properties. UCLA Anderson MSBA ’26.

See my experience → Get in touch
Email LinkedIn
$300K
Application Sales
40%
Cycle Time Cut
88%
NL-to-SQL Accuracy
100K
Records Analyzed
PythonSQLTableauPyTorch · MLFigmaPostgreSQLSnowflakePower BIAgile / Scrum
PythonSQLTableauPyTorch · MLFigmaPostgreSQLSnowflakePower BIAgile / ScrumRoadmappingA/B TestingDatabricksREST APIsJIRA / Confluence
$1.5M
Programs led at Accenture with 17-person teams
35%
Conversion tracking accuracy improvement
60%
Cycle time cut across 6 workflows
18%
Customer base growth from data-driven algorithm
Top 10
Innovation challenge rank across 25,000 global submissions
1.2M AUD
Revenue added through product expansion at TCS
UCLA Anderson MSBAStatistical foundations, ML, and prescriptive models for decision-making. Expected Dec 2026.
9 Years at Fyxt, UCLA, Accenture & TCSFrom SAP ABAP automation to competitive benchmarking and AI product management — delivered measurable business impact.
AI Product Management at FyxtBuilding NL-to-SQL, LLM monitoring and RAG products for property management — 17,000 properties covered.
NFL Big Data Bowl 2026Deep learning on 650K+ tracking records to predict play outcomes before they happen.
01  About Me

Product Manager. Builder.
Storyteller.

I am a Master of Science in Business Analytics (MSBA) student at UCLA Anderson, passionate about blending data, product thinking, and storytelling to drive impact. I specialize in using analytics and machine learning to solve real-world business problems — from predicting player movements in the NFL Big Data Bowl to building a natural-language-to-SQL system at Fyxt.

Before UCLA, I worked on projects that spanned business intelligence, product strategy, and data visualization — building tools that turned complex data into actionable insights. My work bridges the gap between data science and product management: identifying the “why” behind data, designing solutions that scale, and communicating results clearly across teams.

Today I am building AI products at Fyxt Inc. as an AI Product Management Intern, and exploring opportunities where I can apply data-driven decision-making to product development, strategy, or analytics innovation.

02  Skills & Tools

Skills & Tools

Product & Program
Roadmapping
OKRs
RICE / MoSCoW
PRDs
Product Strategy
Stakeholder Management
A/B Testing
Agile / Scrum
Data & Technology
Python
SQL
PostgreSQL
Tableau
Power BI
PyTorch
scikit-learn
Snowflake
Databricks
REST APIs
Figma
JIRA / Confluence
Certifications
SAFe Product Owner / PM 5.1
Microsoft AI-900
IIBA Business Data Analyst / ECBA
CSPO
SAS AI Foundations
INFORMS Business Problem Framing
03  Journey

Experience

AI Product Management Intern ● Current Role

Fyxt Inc.
Los Angeles, CA
Jun 2026 – Present
88% NL-to-SQL accuracy−30% incidents70% adoption in 3 months100K records processed~$163K/yr saved
LLMsNL-to-SQLRAGPostgreSQLA/B TestingFivetran / Airbyte
  • Natural-Language-to-SQL: Built a three-LLM system reaching 88% accuracy, making property data queryable in plain English.
  • LLM Monitoring: Shipped a monitoring tool that cut incidents 30% (20 to 14 per week), saving about $163K/year.
  • AI Scorecard: Launched a scorecard adopted by 70% of users within 3 months.
  • Data Pipeline: Built a PostgreSQL ETL pipeline processing 100K property records; improved data access time from 13 min to 4 min with 88% defect-free delivery.
  • RAG Vendor Engine: Built over 500 documents, adopted by 63 property managers covering 17,000 properties.
  • Experimentation & Discovery: Ran A/B testing across 16 tools with a 5% phased rollout; defined 8 features and 46 user stories across 3 personas; ESG dashboard made reporting 3× faster.

Graduate Research Assistant

University of California, Los Angeles
Los Angeles, CA
Jan 2026 – Present
NLPAIResearchData Analysis
  • AI-Driven Customer Perception Research: Applied NLP, entropy scoring, and AI likelihood scoring to large-scale review datasets. Quantified AI-driven shifts in language authenticity and consumer perception, producing structured research reports delivered to faculty and academic stakeholders.

Business Architecture Associate Manager

Accenture
Gurugram, India
Oct 2022 – Sep 2025
$1.5M programs led$300K first-year sales40% cycle time cut+35% tracking accuracy
PythonTableauSQLPower BIStakeholder Management
  • Programs & Go-To-Market: Led $1.5M programs with 17-person teams; competitive benchmarking across 10+ data sources (Python, Tableau) generated $300K in first-year sales.
  • Executive Reporting: Power BI KPI dashboards cut approval cycle time by 40%; conversion tracking accuracy improved 35%.

Business Architecture Team Lead

Accenture
Gurugram, India
Oct 2020 – Sep 2022
60% cycle time cut · 6 workflowsTop 10 of 25,000 submissions
Process ImprovementJIRAConfluenceBRD / FRD
  • Process Improvement: Process improvement across 6 workflows reduced cycle time by 60%.
  • Innovation Challenge: Intelligent Driver System reached Top 10 of 25,000 global submissions.

IT Analyst

Tata Consultancy Services
Kolkata, India & Melbourne, Australia
Nov 2013 – Feb 2019
+18% customer base1.2M AUD revenue added68K AUD new engagements
SAP ABAPSQLPythonAutomation
  • Data-Driven Product Expansion: Grew the customer base 18% and added ~1.2M AUD in revenue.
  • Workflow Automation: SAP ABAP automation workflows secured 68,000 AUD in new client engagements.
Education

Master of Science in Business Analytics

UCLA Anderson School of Management
Los Angeles, CA
Expected Dec 2026 · GPA 3.8
  • Focused on statistical foundations, data management, and prescriptive models for decision-making.

MBA – Analytics & Marketing

Great Lakes Institute of Management
Chennai, India
2020 · Magna Cum Laude
  • Graduated Magna Cum Laude, specializing in data-driven business analytics and marketing strategy.

BS – Computer Science

Maulana Abul Kalam Azad University of Technology
Kolkata, India
2013
  • Specialized in building algorithms and distributed systems.
04  Writing

Product Management Blog

View all posts →
Product Management

Does an MBA make you a good Product Manager?

Exploring why frameworks matter less than applying them — and how structured thinking turns chaos into clarity.

Nov 2025  ·  4 min read
Product Strategy

Product Management Lessons from a Long-Distance Marriage

A humorous take on prioritization, communication, and roadmap alignment — through the lens of marriage.

Oct 2025  ·  3 min read
Discovery

Always Ask: “Why this Feature?”

How focusing on user intent transforms feature requests into impactful product outcomes.

Sept 2025  ·  5 min read
05  Projects & Research

Selected Projects

★ Featured Project

NFL Big Data Bowl 2026 – Prediction Track

Kaggle Competition | Sports Analytics & Predictive Modeling
650K+ tracking recordsPyTorch deep learningGroupKFold no-leakage validation
Problem

Predict football play outcomes before they occur, using Next Gen Stats player-tracking data — without letting future information leak into the model.

What I Built

Preprocessed 650K+ tracking records and implemented deep neural networks in PyTorch, with GroupKFold cross-validation to prevent data leakage.

Outcome

A leakage-free predictive pipeline that forecasts play outcomes pre-snap, validated on grouped splits of real NFL tracking data.

PyTorchDeep LearningGroupKFoldSports Analytics

AI-Powered Content Moderation System

Machine Learning & Content Safety
92% harmful detection96% spam detection<70ms latency
Problem

Harmful, spam, and threat content needs to be caught automatically at scale — fast enough not to slow the product down, with humans still in control of edge cases.

What I Built

An end-to-end ML moderation pipeline in Python: a multi-label text classifier using scikit-learn and NLP feature engineering, combined with a human-in-the-loop review architecture.

Outcome

92% harmful and 96% spam detection at under 70ms per item.

NLPClassificationscikit-learnHuman-in-the-loop

AI-Enabled Feedback & Semantic Search Platform

Serverless AI & Vector Search
1K+ feedback records<100ms query latencyReal-time semantic matching
Problem

Manual keyword searches could not surface the right customer feedback — the words people use rarely match the words in the records.

What I Built

A fully serverless semantic search platform on Cloudflare Workers and Vectorize, using embeddings to ingest and query 1K+ customer feedback records in real time.

Outcome

AI-driven semantic matching replaced manual keyword searches, answering queries in under 100ms.

Cloudflare WorkersVectorizeEmbeddingsSemantic Search

Food Agent – AI-Powered Restaurant Recommendation System

Full-Stack Web Application & AI
Geo-aware rankingNatural language queriesReal-time results
Problem

People describe what they want in plain language — “cheap breakfast open now within 2 miles” — which rigid filter forms handle poorly.

What I Built

A full-stack, geo-aware recommendation system (Next.js + FastAPI) that parses natural-language queries and ranks nearby options by distance using the Haversine formula, powered by the Google Places API.

Outcome

Nearby restaurants returned in real time with ratings, distance, open/closed status, and direct Google Maps links.

Next.jsFastAPIGoogle Places APIHaversine
06  Get in Touch

Let’s Connect

If you would like to collaborate, network, or just say hi — I would love to hear from you. I am open to conversations about product management, technical program management and AI product roles.

Email LinkedIn 213-809-6966