Open to opportunities

Sanket Bhattacharjee

@sanketbhattacharjee

Message

Data Scientist focused on applied AI—multimodal vision-language models, LLM adaptation, and production-ready pipelines.

Zimbabwe

Message

What I'm looking for

I’m open to India, Europe, and global remote roles, and I’m looking for opportunities where I can build multimodal and generative AI systems end to end—model adaptation, validation, and reliable data/production pipelines.

I’m a Data Scientist at Fractal Analytics, working across multimodal AI and vision-language models, with hands-on experience in parameter-efficient fine-tuning and production AI workflows.

In a multimodal chest X-ray project, I built batch inference pipelines using a locally deployed MiniCPM vision-language model, then moved from weak baselines to LoRA/PEFT fine-tuning on frontal and lateral X-rays with multi-label annotations across approximately 13–15 conditions.

I also contribute to enterprise AI by engineering and validating data pipelines for an internal GPT application (PySpark, Databricks, scheduling/orchestration, and source-to-processed-to-portal reconciliation), and I support agent-assisted migration discovery using GitHub Copilot/VS Code with Databricks CLI and Azure CLI workflows.

Experience

Work history, roles, and key accomplishments

Current

Associate, Data Science

Current

Fractal Analytics

Aug 2025 - Present (11 months)

Built batch inference workflows using a locally deployed MiniCPM vision-language model for multimodal chest X-ray disease/finding prediction. Fine-tuned the model with LoRA/PEFT on frontal and lateral X-rays using multi-label annotations and evaluated baseline vs. fine-tuned performance.

Python MiniCPM Vision Language Models Multimodal AI Computer Vision Generative AI Hugging Face LoRAs PEFT Batch Inference Prompt and Inference Configuration Model Evaluation

Enterprise Data Engineering & Validation

Mars GPT

Contributed to the data engineering layer for an internal enterprise GPT application serving financial, sales, KPI, and business-performance information. Performed source-to-processed-to-portal reconciliation checks to validate downstream reliability.

PySpark Databricks Data Engineering ETL Data Pipelines Data Cleaning Filtering Scheduling Orchestration Data Validation Reconciliations

Agent-Assisted Migration Discovery

Mars GPT

Supported migration discovery using reusable Markdown-based AI agent instructions in GitHub Copilot/VS Code with Databricks CLI and Azure CLI workflows. Analyzed pipeline and job metadata to identify runtime variation and migration-relevant operational inconsistencies, with exposure to ADF and cloud/data-platform metadata workflows.

GitHub Co Pilot AI Agents Process Metadata Operational Discovery

Multimodal Video Analytics

Mars GPT

Contributed to cross-environment replication of a Qwen-based video analytics workflow that interpreted cat-feeding behavior and passed inferred signals to downstream classification for preference assessment. Worked on environment compatibility, configuration/path adaptation, dependency handling, and workflow validation.

Qwen Multimodal AI Video Analytics Configuration and Path Adaptation Validation