Varun Kumar@Vakum
2
Events attended
2
Submissions made
India
5+ years of experience
About me
I am a Consultant with a strong foundation in financial systems, enterprise architecture, and next-generation AI technologies. Currently working at Deloitte, I contribute to SAP-focused projects, primarily in Identity & Access Management (IAM) and cloud-based AI solutions. Over the course of my career, I have built and optimized mission-critical systems, including trading engines, wealth management platforms, and access management solutions. At InvestCloud, I worked on key products like the Unified Wealth Platform (UWP) and Trade Network, enabling large-scale trading operations and reporting. At Deloitte, I have expanded into Generative AI, building enterprise-grade solutions such as a SAP BTP chatbot using RAG and LLM embeddings and automation tools for IAM workflows. My technical expertise includes Java, Python, SAP BTP, CAP, and SQL, supported by certifications in Generative AI and Data Science. I am passionate about bridging enterprise systems with modern AI innovations to drive intelligent automation and scalable digital transformation.
🤓 Latest Submissions
DevBand: Multi-Agent Software Delivery
DevBand is a Band-powered multi-agent software delivery workflow for enterprise engineering teams. Instead of using one isolated coding assistant, DevBand coordinates specialist agents through a shared Band collaboration room. A user submits a software feature request, and DevBand creates a live Band room, adds specialist Band agents as participants, and runs the request through planning, implementation, testing, review, revision, and documentation. The Product Planner Agent creates user stories, acceptance criteria, technical tasks, risks, and handoff context. The Software Engineer Agent generates implementation guidance and code snippets. The Test Engineer Agent creates unit test ideas, integration checks, edge cases, and manual QA steps. The Code Reviewer Agent evaluates quality, maintainability, accessibility, and security. The Software Engineer Revision Agent resolves reviewer feedback. The Documentation Agent generates a final PR-ready package with release notes, developer notes, and README updates. Band is used as the core collaboration layer. DevBand creates real Band rooms, adds six specialist Band agents as participants, and records each stage as structured Band task events with agent roles, handles, handoffs, and completion status. AI/ML API powers the agent reasoning and generation, while the frontend visualizes the collaboration timeline, workflow metrics, Band participants, agent outputs, and final delivery package. The final output is a downloadable Markdown PR package that provides an audit-ready software delivery artifact. DevBand demonstrates how enterprise software delivery can move beyond single-agent code generation into traceable, collaborative, multi-agent workflows.
19 Jun 2026
VendorWatch AI
Enterprises rely on hundreds or thousands of third-party vendors, SaaS tools, outsourcing partners, suppliers, and infrastructure providers. However, vendor risk signals often appear first across the public web before they reach internal risk systems. These signals may include cybersecurity incidents, data breaches, lawsuits, regulatory investigations, service outages, layoffs, financial instability, leadership changes, customer backlash, and reputational issues. VendorWatch AI solves this problem by acting as an autonomous vendor-risk analyst. Given a vendor name and region, the agent plans a risk investigation, generates targeted risk-specific search queries, and uses Bright Data MCP to retrieve live public web intelligence. The system then extracts structured evidence, classifies findings into enterprise risk categories, calculates an explainable risk score, and generates an enterprise-ready Markdown and JSON risk report. The project is designed for procurement, compliance, security, finance, and third-party risk teams that need faster and more evidence-backed vendor monitoring. Instead of relying only on periodic questionnaires or manual research, VendorWatch AI helps teams move toward proactive, live web-based risk intelligence. The current MVP includes a Streamlit dashboard, Bright Data Remote MCP integration, live web investigation through the `search_engine` MCP tool, agent execution trace, structured evidence table, risk score breakdown, downloadable Markdown report, downloadable JSON output, and transparent Bright Data tool usage. VendorWatch AI demonstrates how AI agents can use production-grade web data infrastructure to solve real enterprise risk and compliance problems.
31 May 2026
