Göteborg, Sweden

Behavioral modelling,
chess & AI research

I'm Ravinder Reddy Atla — a solo ML researcher building a personal chess AI lab on top of 10,000+ personal games and the Lichess elite database. My work sits at the intersection of behavioral modelling, human-AI alignment, and skill progression — using chess as a model system where ground truth exists and data is abundant.

I'm particularly interested in what makes human decision-making at different skill levels measurably distinct, and whether those differences can be used to build genuinely personalised coaching systems — not just engine analysis with a friendlier UI.

Current focus — Chess AI Lab

April 2026

Active projects · C1 → R3

C1
Behavioral fingerprint
Longitudinal style tracking from 10K personal games
C2
Mistake coach
Blunder classifier + GM context retrieval
C3
Repertoire builder
Style-aware opening recommendations
C4
Opponent profiling
Pre-game brief from public game history
C8
Clock agent
Time-pressure behavioral dataset
R1
Rating band atlas
What changes behaviorally per 100 Elo
R2
GM curriculum
Style-matched GM study plan generator
R3
Fair-play dataset
Labeled behavioral dataset for detection research
Chess AI Lab v2 — full project map
8 projects spanning behavioral fingerprinting, RL-based coaching, personalised repertoire building, opponent profiling, and research-grade atlas construction. Each project has an abstract, research questions, ordered task list, and AI/ML concept tags.
Behavioral modelling Human-AI alignment Skill progression CSSLab adjacent Lichess · chess.com LangGraph · RAG · RL

Research interests

Behavioral modelling

Individual decision-making signatures — how move distributions, time allocation, and positional preferences form stable, identifiable fingerprints across time.

Skill progression

What measurably changes in how people think and decide as skill increases — building a behavioral atlas of improvement, not just an Elo trajectory.

Style-matched learning

Personalised coaching that accounts for who you are, not just what the engine recommends — matching study material to your specific behavioral profile.

Human-AI alignment

Chess as a model system for studying how humans learn from and collaborate with superhuman AI — and how AI can be made genuinely useful to humans, not just optimal.

Writing & notes

in progress

First post coming after C1 ships.