GrepBio multiBoard

A cloud-native platform for automated bioinformatics analysis and interactive scientific visualisation

multiBoard is GrepBio's SaaS platform for automating complex bioinformatics analysis workflows. I co-founded GrepBio with Dr. Kalaivani Raju and serve as its CTO, with responsibility for the overall technology strategy, software architecture, cloud infrastructure, and engineering of the platform.

The goal behind multiBoard was to turn traditionally complex, command-line-driven bioinformatics workflows into a reproducible and accessible web platform. Researchers can submit datasets, execute computationally intensive analysis pipelines on cloud HPC infrastructure, explore the resulting biological data through interactive visualisations, and receive automatically generated scientific reports.

Platform Architecture

I designed the overall architecture and technology stack for multiBoard, spanning the web application, backend services, computational pipelines, cloud infrastructure, and reporting systems.

The platform consists of several major components:

• React/TypeScript frontend provides the user-facing dashboard for dataset submission, workflow configuration, analysis monitoring, and interactive exploration of results.
• FastAPI/Python backend exposes the REST API used by the frontend and handles authentication, application logic, data management, and orchestration of computational workflows.
• MongoDB serves as the primary application database and metadata store for analyses and their results.
• Snakemake-based bioinformatics engine combines Python, R, shell scripts, and established command-line bioinformatics tools into reproducible computational workflows.
• Slurm-managed HPC infrastructure in Azure executes the computationally intensive bioinformatics pipelines in parallel.
• File-system storage holds the datasets and analysis artefacts exchanged between the bioinformatics pipelines and downstream components.
• Automated scientific reporting engine uses Python, Jinja2, and LaTeX to transform analysis results and generated artefacts into structured PDF reports.
• Dolphin, a shared database-schema and data-model package implemented using Python and Pydantic, defines the typed data contracts used across the backend, analysis, storage, and reporting components.

The architecture deliberately separates the user-facing application, persistence layer, computational workflows, file storage, and report generation. These components communicate through well-defined REST and typed data interfaces, allowing each part of the platform to evolve independently while supporting increasingly complex analysis workloads.

Building a Modular Analysis Platform

multiBoard initially focused on a smaller set of genomic analysis workflows, but customer requirements quickly expanded the scope of the platform. Working closely with our bioinformatics team, I evolved the system into a modular analysis architecture capable of supporting multiple independent workflows, including:

• Bulk RNA sequencing
• Single-cell RNA sequencing
• Amplicon sequencing
• Metagenomics
• Metatranscriptomics

Each workflow presented different computational, data-processing, and visualisation requirements. Rather than treating them as separate applications, we extended the common platform architecture so that new analysis modules could reuse the same infrastructure for job orchestration, data management, execution monitoring, result storage, reporting, and user interaction.

Interactive Biological Data Visualisation

A significant engineering challenge was presenting very large and highly interconnected biological datasets in a way that researchers could explore interactively. We developed custom visualisation components that combined primary analysis results with external biological knowledge, including:

• protein-protein interaction networks,
• biological process and pathway annotations,
• cellular component annotations,
• disease and cancer associations, and
• functional annotations.

Instead of producing only static analysis outputs, multiBoard allowed researchers to navigate relationships between genes, proteins, pathways, diseases, and biological functions directly within the application. This required designing frontend components capable of representing complex graph and multidimensional datasets while maintaining a usable interactive experience.

Automated Scientific Report Generation

In addition to the interactive dashboard, I designed a separate reporting engine for producing publication- and review-ready PDF reports automatically. The reporting service retrieves completed analysis results and metadata from MongoDB and the HPC FileSystem Storage, transforms them into structured report models, and renders the final documents using Jinja2-driven LaTeX templates.

This architecture enabled us to generate detailed reports containing analysis summaries, figures, tables, statistical results, and biological interpretations without requiring researchers or bioinformaticians to manually assemble documents after each analysis. The reporting engine was deliberately kept independent from the web application and analysis pipelines, allowing report templates and presentation logic to evolve without affecting computational workflows.

Cloud and Infrastructure

I was responsible for designing and operating the Azure infrastructure supporting the platform, including the application services and the HPC environment required by the analysis engine.

GrepBio was selected for the Microsoft for Startups program and received USD 50,000 in Azure cloud credits, which enabled us to build and operate the platform's initial production infrastructure.

We subsequently received a Startup India grant of about USD 43,000, alongside our own bootstrapped investment, to continue developing the platform and expanding its analytical capabilities.

As customer adoption increased, the architecture had to accommodate growing datasets, longer-running computational workloads, additional analysis pipelines, and increasingly sophisticated visualisations while keeping the user-facing application isolated from the complexity of the underlying HPC environment.

multiBoard ultimately became much more than an interface around bioinformatics scripts. It evolved into an integrated scientific computing platform combining SaaS application architecture, workflow orchestration, HPC computing, large-scale data processing, interactive visualisation, and automated document generation within a single system.