I'm a Senior AI/ML Engineer at Nielsen, specializing in Data Science and Machine Learning across multiple domains. I build practical, production-ready solutions that drive real-world impact, play with data and try telling interesting stories through the data, research and learn about new technologies and develop new techniques to solve probles

Classical Machine Learning

• Time series prediction and forecasting
• Clustering and Basic Statistics modelling like XGBoosting, LightGBM, KNN clustering, Random Forest and all
• Time series and Prophet Modelling

Computer Vision

• Visual Cryptography implementation
• Hash matching for video frame identification and analysis
• Model fine-tuning: YOLO-v8, ResNet-50, ResNet-200

Natural Language Processing & LLMs

• NLP applications and solutions
• Large Language Model (LLM) implementations
• Retrieval-Augmented Generation (RAG) systems

Emerging AI Technologies

• Agentic AI development and deployment
• GenAI integration
• Model fine-tuning: Gemma, MiniLM, GPT
• Advanced prompt engineering and LLM orchestration
• KnowledgeGraph

Tools

• FAISS, vectorDB

• graph database

• Tableau

• Cloud: AWS, GCP

• Open-source research and documentation

LatentGuard — Dynamically Steering Multilingual Safety Boundaries

An ongoing AI safety research project investigating whether safety behaviours learned primarily from English generalize reliably to Bengali.

The work explores:

• Behavioural safety evaluation
• English–Bengali safety comparisons
• Token fragmentation and refusal mismatches
• Hidden-state and representation analysis
• Cross-lingual representation drift
• Latent steering and representation-level interventions

The project was selected among the Top 10 finalists of the Global South AI Safety Hackathon.

Research code and accompanying work are being developed openly.

Multilingual Representation Learning — MRL 2026

Actively contributing to the 2026 Multilingual Representation Learning benchmark, including Bengali-English language data, with a focus on improving evaluation coverage for underrepresented languages.

Natural Language Inference (NLI) Research

• Investigated dataset artifacts and spurious correlations by fine-tuning ELECTRA-small on the SNLI dataset.
• Designed interventions using expanded datasets to mitigate misclassification patterns and improve generalization.

Deep & Reinforcement Learning Systems

• Programmed autonomous vision-based and state-based Reinforcement Learning agents to navigate complex, real-time adversarial dynamics, such as automated SuperTuxKart competitive environments.

• Senior AI/ML Engineer at Nielsen, driving innovation in data science and machine learning
• Research-focused professional actively contributing to company research initiatives
• Actively exploring AI Safety, Multilingual NLP, Representation Learning and LLM research
• Passionate about staying at the forefront of AI/ML advancements
• Experienced in deploying production-grade ML solutions
• Interested in understanding model behaviour, evaluating failure modes, and developing robust ML systems
• Open to collaborating on research projects, publications, open-source research, and innovative ML solutions

🌱 Open to collaborating on research projects, publications, and innovative ML solutions!