Jatin Roy, known online as ajeebtech, is a developer and machine-learning researcher who builds consumer applications and tools. His work includes VODs Space for VALORANT match analysis, computer-vision tooling and research on financial and ESG time series.[1][2][3]
Start with Vods Space →
X Career
Roy’s portfolio records an ML research internship at IIT Kharagpur from May to December 2025, founding the scrimmy project from March to May 2026, and a product-engineering contract at morat.ai from June to August 2026. His published work connects research experiments with applications for reviewing esports footage.[1]
Contributions
Turn match recordings into round-level evidence
VODs Space helps viewers find VALORANT recordings and review play. Roy’s round-tracking repository explains how its pipeline samples cropped timer frames, uses a vision model to read them and turns the result into round and clip metadata. This makes a long recording navigable by the events a player or coach wants to study.[4][5]
Preserve meaning across heterogeneous time series
Roy co-authored Learning the ESG Geometry with Domain Aware Language Models, presented at ICML 2026. The paper combines ESG indicators, financial returns, news and sentiment using value-aware tokens and orthogonal embeddings. Its geometry-preserving objective and FACET tokens represent temporal trajectories for forecasting and grouping entities.[3]
Tools, writing & talks
Apps
Start here
Vods Spacevods.spaceA VALORANT recording and replay site for finding matches and reviewing gameplay.
Choose a familiar match and inspect how the replays are organized. Then read the linked round-tracker repository to see how timer recognition becomes clip metadata; it provides a concrete computer-vision workflow behind the interface.
Creator[1][5]Original source
Tools
- VALORANT round tracker
github.comA vision-assisted pipeline for extracting round timing and clip metadata from match recordings.
Developer[4]
Writing, research & talks
ICML 2026 research on representing numerical, semantic and temporal relationships across financial and ESG signals.