Vishal Sharma
Data & AI — writing, building, and sharing in the open
I write and build in the open about Data, AI, GenAI, Agentic AI, and analytics — exploring how these systems actually work in practice, sharing what I learn, and answering questions in the open.
Axiom Macro
Autonomous Macroeconomic Intelligence Terminal
My newest build — a live, open research terminal that reads global recession risk and where the market cycle stands, in real time. The quant is deterministic Python; the AI only narrates. Open access, no login.
- Market-cycle stage engine
- Recession odds across ~15 economies
- Composite systemic-risk score
- Nightly grounded AI briefing
About
Thinking out loud about Data & AI
I work in Data & AI and use this space to think out loud — writing about GenAI, Agentic AI, RAG, BI, and cloud analytics, and building small open-source demos to explore the ideas hands-on.
Core Capabilities
Capability Areas
Generative AI & RAG
Secure assistants, document chat, semantic retrieval, and LLM automation.
Agentic AI Automation
Autonomous workflows for finance, operations, and approvals with human-in-the-loop.
Data Engineering
ETL/ELT, data lakes, cloud platforms, and pipeline modernization.
BI & Executive Analytics
Power BI dashboards, KPI reporting, and embedded analytics.
Cloud Analytics
Azure, AWS, and GCP data platforms built for scale and governance.
AI Strategy & Readiness
Readiness assessment, roadmaps, governance, and executive advisory.
Open source
Demos You Can Try
Small, free, open-source demos I've built to explore ideas hands-on. Try them in the browser or run them yourself.
Document Intelligence — Ask Your Files
A live, gated demo: upload a PDF or text-based document and ask questions in natural language. LLM-agnostic — the same app runs on local models (Ollama, Llama) or any cloud LLM. This instance is powered by Gemini.
Structured Extraction — Documents to Clean Data
A live, gated demo: upload an invoice, receipt, résumé, contract, or business card and get back clean, schema-aligned JSON — fields, line items, and a one-line summary. LLM-agnostic; this instance uses Gemini.
Talk to Your Data — Natural-Language Analytics
A live, gated demo: load a CSV (or a built-in sample) and ask questions in plain English. The AI writes SQL, runs it against an in-memory SQLite engine, and returns a results table plus an auto-generated chart. LLM-agnostic; this instance uses Gemini.
Chat with the Articles — Grounded RAG Assistant
A live, gated demo: ask questions about Vishal’s published articles and get answers grounded strictly in them, with citations linking back to each post. The same RAG pattern the blog explains, running live. LLM-agnostic; this instance uses Gemini.
Auto-EDA Dashboard — Instant Data Profiling
A live, gated demo: upload a CSV (or a built-in sample) and get an instant exploratory data analysis — column profiling, auto-generated charts, and an AI-written summary of what stands out. LLM-agnostic; this instance uses Gemini.
Data Quality Scorecard — Trust Your Data
A live, gated demo: upload a CSV and scan it for missing values, duplicates, statistical outliers, and inconsistent formats. Get a per-column and overall quality score with severity-rated issues and AI-prioritized fixes. LLM-agnostic; this instance uses Gemini.
CSV Cleaner & Transformer — Wrangle in Plain English
A live, gated demo: load a CSV, then describe a cleanup in plain English — trim, standardize, split columns, deduplicate, filter. The AI writes a safe read-only transformation, shows a preview, and lets you download the cleaned file. LLM-agnostic; this instance uses Gemini.
Data Dictionary Generator — Document & Govern
A live, gated demo: upload a CSV and auto-generate a governance-ready data dictionary — inferred business meaning per column, PII flags, and sensitivity ratings — downloadable as Markdown. LLM-agnostic; this instance uses Gemini.
Interactive Multimodal Playground — See & Parse Any Image
A live, gated demo: drop in an image — a receipt, chart, photo, or screenshot — and the AI parses it on the spot. Describe the scene, detect objects, read text (OCR), extract data from a chart or table, or ask your own question. Start instantly with a sample. LLM-agnostic; this instance uses Gemini.
Experience
18+ years across Data, AI, Cloud & Enterprise Delivery
Data & AI Strategy
How enterprises assess data and AI readiness, spot real adoption opportunities, build roadmaps, and put sensible governance in place — the strategy side I think and write about.
Generative AI and RAG Solutions
Design and development of secure AI assistants, document chat systems, enterprise knowledge search, semantic retrieval, and LLM-powered automation.
Agentic AI Workflow Automation
AI agents for business process automation, finance operations, vendor queries, document processing, approvals, exception handling, and human-in-the-loop workflows.
Power BI and Executive Analytics
Executive dashboards, finance dashboards, operational reporting, KPI reporting, embedded analytics, and business intelligence modernization.
Data Engineering and Cloud Platforms
ETL/ELT pipelines, data lake architecture, cloud data platforms, Azure Data Factory, SQL optimization, API integration, and data platform modernization.
AI Delivery and Solution Architecture
End-to-end solution design, architecture review, effort estimation, delivery planning, stakeholder alignment, and proof-of-value execution.
Insights
Recent Articles
Background Agents Arrived — and Brought Rationing With Them
In three weeks, asynchronous agents crossed from feature to platform: background by default, drivable from a phone, governed by admins. In the same releases, vendors started capping how many an agent may spawn. Both halves of that sentence matter.
The Agent Statistics Everyone Quotes Are a Year Old
The "95% of AI pilots fail" figure is from July 2025. The "40% of agentic projects will be cancelled" line is a June 2025 prediction about 2027. If your 2026 business case rests on either, it rests on evidence collected before most of the tools in it existed.
Agent Security Is Being Patched, Not Solved
In three weeks: a GitHub agent tricked into publishing private repositories by an unauthenticated stranger, a coding assistant that executed untrusted config on repo open, and a run of permission-bypass fixes in a major agent CLI. The pattern is consistent, and it is architectural.
Let's talk technology
Have a question about Data & AI, or an idea you want a second opinion on? Ask in the open forum — it's free, and I read every thread.