Data Scientist (Generative AI, Bayesian Inference)

About the Company and Project

ARC-Net supports deep-tech projects in taking the critical 0-to-1 step, specializing in foundational research and applied AI. Incubated at ARTPARK, Indian Institute of Science (IISc), ARC-Net provides scientific, technical, and engineering capability for emerging technologies.

You will be working on MyPraX, a cognitive-AI platform that models how people perceive and interpret advertising to predict real brand uplift. The platform integrates multimodal video analysis, neuro-semantic token mapping, Bayesian inference, and temporal cognition modelling. MyPraX is currently building its first working demo and v0 prototype, which will become the foundation of a full-fledged product for creative pre-testing and brand development.

The role is to develop and optimise a multimodal AI system that extracts neuro-semantic tokens from advertising content and predicts brand uplift using Bayesian inference and time-series modelling.

Contract : Project-based (with potential long-term extension)

Compensation : ₹70,000 / month (with potential increase as the project scales)

Location : On-site, Bengaluru

Start Date : Immediate

Core Responsibilities

Build and maintain pipelines for multimodal encoding(text, audio, video) using models such as SBERT, AST, and ViTs.

Use Neural networks to create high-resolution spatio-temporal heatmaps over videos.

Develop Bayesian inference frameworks.

Create time-series forecasting models.

Preferred (Good-to-Have) Skills

Experience with attention-based multimodal fusion architectures.

Familiarity with generative AI for content interpretation.

Understanding of causal inference methods.

Former experience with marketing and / or neuroscientific data analysis.

Education / Qualifications

Bachelor’s or Master’s degree in Computer Science, Data Science, Applied Mathematics, Statistics, Computational Neuroscience, or a related field.

PhD preferred but not mandatory if backed by strong applied research experience.