Football Research Scientist

Rishav
Dutta.

Where the gridiron meets the algorithm.
Currently at the Cleveland Browns.
Rishav Dutta
47.6062°N // 122.3321°W
Scroll
01 — About

The Combine.

Senior AI Research Scientist with the Cleveland Browns, operating at the intersection of machine learning and professional football. Four NFL seasons of building models that quantify what scouts feel and coaches know.

Trained at Carnegie Mellon University's School of Computer Science — B.S. in Computer Science with a Minor in Machine Learning, supplemented by research in Statistics & Data Science. Six peer-reviewed publications spanning computer vision, NLP, and sports analytics.

EducationCarnegie Mellon SCS — B.S. CS, ML Minor
RoleSr. AI Research Scientist, Cleveland Browns
Seasons4 NFL seasons
Published6 publications — JQAS, Big Data Bowl, CMU
Off the field

Plays football, soccer, cricket, and basketball. Browns faithful since day one, Seattle Sounders supporter. Shares the apartment with two cats who couldn't care less about any of the above.

Combine Results
Draft GradeA+
MeasurableGrade
Cleveland, OH2021 – Present
Live
02 — Projects

The Film Room

Breaking down the game tape. Selected research, publications, and tools — reviewed frame by frame.

Rec
Play
CH-02
11 Clips 00:00:00:00
01
PaVE
PaVE: Passing Value in Expectation2021
PublicationCNN
REC
Publication2021
PyTorch
Keras
CNN
Kaggle // NFL Big Data Bowl 2021
TC 01:12:04:18
Scouting Report — Frame 01
PaVE the Way for
NFL Passing Analytics
Situation
NFL Big Data Bowl 2021 entry. Quantifying what passing value actually looks like using spatial deep learning on tracking data.
Key Reads
  • CNN-based spatial modeling of the football field with zone-level predictive value
  • Quantified defensive impact and QB quality through player tracking data
  • Assigned credit to individual players via model input analysis
Result
Honorable Mention — NFL Big Data Bowl
Personnel
Ranasaria, Dutta, Sanjeev, Murali
Stack
PyTorch, Keras, CNN
View Full Breakdown
02
NFL3D
NFL3D: Adding the Third Dimension2021
ResearchR
NFL3D
REC
Research2021
R
3D Viz
CMU Sports Analytics Conference
TC 02:34:11:06
Scouting Report — Frame 02
NFL3D: Adding the
Third Dimension
Situation
NFL tracking data is 2D. The ball moves in 3D. That gap was the problem.
Key Reads
  • Algorithm to interpolate z-axis from publicly available tracking data
  • Enables precise 3D ball trajectory analysis on any play
  • Real-time flight path estimation at any point during a play
Result
Conference Presentation — Research Poster
Personnel
Dutta, R. (2021)
Stack
R, Interpolation
View Full Breakdown
03
nflscrapR
nflscrapR2020
R PackageOpen Source
nflscrapR
REC
R Package2020
R
EPA/WPA
Recognized by NFL Dir. of Data & Analytics
TC 03:07:22:09
Scouting Report — Frame 03
nflscrapR: NFL
Play-by-Play in R
Situation
The football analytics community needed clean, accessible play-by-play data. We built the pipeline.
Key Reads
  • R package for seamless retrieval of NFL play-by-play data via web scraping
  • Expected Points (EPA) and Win Probability (WPA) models built in
  • Adopted by The Athletic, FiveThirtyEight, and the broader analytics community
Result
Industry Standard — Used by Major Outlets
Personnel
Horowitz, Yurko, Ventura, Dutta
Stack
R, Web Scraping, API
View Full Breakdown
04
Unsupervised DBs
Unsupervised Methods for Pass Coverage2019
JournalML
Unsupervised DBs
REC
Journal Paper2019
Python
GMM
Journal of Quantitative Analysis in Sports
TC 04:19:33:21
Scouting Report — Frame 04
Unsupervised Methods for
Identifying Pass Coverage
Situation
Man or zone? The tracking data knows, but nobody had taught a machine to see it without labels.
Key Reads
  • Unsupervised model to classify man vs zone from player motion data
  • Gaussian Mixture Models + hierarchical clustering on constructed features
  • Advised by Samuel Ventura and Ronald Yurko at CMU
Result
Published — JQAS
Personnel
Dutta, Yurko, Ventura
Stack
Python, Clustering, PCA
View Full Breakdown
05
NGS
NGS Play Animation2019
ToolData Viz
NGS
REC
Tool2019
Python
Matplotlib
NFL Next Gen Stats Tracking Data
TC 05:42:08:14
Scouting Report — Frame 05
Next Gen Stats
Play Animation
Situation
Tracking data is rows in a spreadsheet. This tool turns it into film you can actually watch.
Key Reads
  • Animates all 22 players from NGS tracking data in real time
  • Frame-by-frame rendering of any play in the dataset
  • Visual tool for pattern recognition and play breakdown
Result
Analysis Tool — Internal Use
Personnel
Dutta, R.
Stack
Python, Matplotlib
View Full Breakdown
06
Game Theory
Game Theory Approach to Basketball2019
ResearchNBA
Game Theory
REC
Research2019
Python
Nash Eq
CMU Sports Analytics Conference
TC 06:15:47:02
Scouting Report — Frame 06
A Game Theoretic
Approach to Basketball
Situation
Every possession is a strategic game. We modeled it as one.
Key Reads
  • Modeled NBA possessions as one-shot games with real matchup data
  • Nash Equilibrium to identify optimal mixed strategies per team
  • Predicted the most-used strategy for 2020 NBA Finals before the series
Result
Predicted 2020 NBA Finals Strategy
Personnel
Dutta, Chakrabarti
Stack
Python, NBA PBP Data
View Full Breakdown
07
Sofy.AI
Sofy.AI: AI-Driven Web Testing2017-20
ProductML
Sofy.AI
REC
Product2017–2020
Python
Azure
Lead ML Developer // Sofy.AI
TC 07:28:55:11
Scouting Report — Frame 07
Sofy.AI: AI-Driven
Web Testing Platform
Situation
Manual web testing is slow and misses bugs. We built an AI to explore and break websites automatically.
Key Reads
  • Led cross-functional team of engineers and data scientists
  • ML systems using graph theory to optimize test coverage and find defects faster
  • Cloud architecture on Azure with CI/CD and high availability
Result
Shipped Product — 3 Year Tenure
Role
Lead ML Developer
Stack
Python, Azure, Graph Theory
View Full Breakdown
08
ICU
Predicting ICU Patient Mortality2018
CompetitionNeural Nets
ICU
REC
Competition2018
Python
NN
PhysioNet Computing in Cardiology
TC 08:51:16:27
Scouting Report — Frame 08
Predicting ICU
Patient Mortality
Situation
PhysioNet Challenge. 48 hours of ICU data. Predict who survives.
Key Reads
  • Quorum of Neural Networks with randomized training and oversampling
  • Thresholded voting scheme across the ensemble
  • Used medical journal averages to handle missing clinical data
Result
Top Score — PhysioNet Challenge
Personnel
Dutta, R. (2018)
Stack
Python, Neural Networks
View Full Breakdown
09
InteLLect
InteLLect2018
Hackathon
InteLLect
REC
Hackathon2018
AI
EdTech
CMU Hackathon // Team of 4
TC 09:03:28:15
Scouting Report — Frame 09
InteLLect
Situation
36-hour hackathon. Build something that makes learning smarter. Clock's ticking.
Key Reads
  • Intelligent systems to personalize educational experiences
  • Built under competitive time constraints with a team of 4
  • Focus on practical, real-world applicability
Result
Completed — 36 Hour Build
Event
CMU Hackathon
Stack
AI, Education Tech
View Full Breakdown
10
Omaha
Omaha: Football Analysis Tool2016
SportsPython
Omaha
REC
Tool2016
Python
Personal Project // Football Analytics
TC 10:22:41:03
Scouting Report — Frame 10
Omaha: Football
Analysis Toolkit
Situation
Before the Browns, before CMU research — this is where it started. A toolkit to break down football the way I saw it.
Key Reads
  • Python pipeline for game data analysis and visualization
  • Automated report generation from play-by-play data
  • Built to uncover strategic insights others weren't looking for
Result
Foundation — Where It All Started
Personnel
Dutta, R.
Stack
Python
View Full Breakdown
11
AI Advent Calendar
AI Advent Calendar: Talking LED Jeopardy2025
CreativeAIHardware
AI Advent Calendar LED Grid
REC
Creative Tech2025
Claude API
Arduino
Hardware + AI // Personal Project
TC 11:25:12:07
Scouting Report — Frame 11
AI Advent Calendar:
Talking LED Jeopardy
Situation
Build a Christmas gift that's never been built before. An advent calendar that talks, plays Jeopardy, and lights up the room — literally.
Key Reads
  • Arduino-controlled addressable LED strip grid functioning as a giant pixel display
  • Claude API and OpenAI API generating dynamic Jeopardy clues and categories
  • AI text-to-speech as the live game show host
  • Full interactive game with real-time scoring and daily reveals
Result
Built & Shipped — Hardware + AI
Personnel
Dutta, R.
Stack
Claude API, OpenAI, Arduino, LED, TTS
View Full Breakdown
Rishav Dutta
Senior AI Research Scientist
Cleveland Browns
End of Reel
03 — Career

The Depth Chart

Every team needs a roster. Here's mine.

RD
Active Roster
2016 — PRESENT
QB1
The Playmaker
Research & Analytics
#1
★ Captain
Senior AI Research Scientist
Cleveland Browns
2021 — Present
◆LLMs for automated scouting reports & player evaluation
◆CNNs for player tracking data & multimodal projections
◆#1 ranked analytics department in the NFL, 2024
LLMsCNNsComputer VisionPythonPyTorch
LT
The Protector
Engineering
#76
Starter
Software Engineer
Microsoft
2020 — 2021
◆Azure Networking — backbone infrastructure
◆10x improvement in allocation speed via GRPC microservices
C#GRPCAzureMicroservices
WR1
The Deep Threat
Machine Learning
#11
Starter
Lead ML Developer
Sofy.AI
2017 — 2020
◆AI-driven web testing platform
◆Graph theory algorithms for intelligent test generation
PythonGraph TheoryAzureNLP
ST
The Specialists
Publications
2021PaVE — Pass Value EstimationHonorable MentionNFL Big Data Bowl
2020nflscrapR — Open-source NFL data toolkitFeaturedThe Athletic / 538
2019Unsupervised Methods for Pass CoverageJQAS
2019NFL3D & Game Theory in BasketballCMU Sports Analytics
2018ICU Patient Mortality PredictionTop ScorePhysioNet
HC
The Coach
Education
2016 — 2020
Carnegie Mellon University
B.S. Computer Science — Machine Learning Minor
School of Computer Science • Statistics & Data Science Research
Impact Over Time

Career WAR

Wins Above Replacement

Current WAR
0
Impact Rating
CMU (2016–2020)
Microsoft (2020–2021)
Browns (2021–Present)
Milestone
04 — Contact

Let's Connect

Always open to conversations about sports analytics, data science, or creative collaborations.

rishavdutta.wa@gmail.com
LinkedIn GitHub Twitter