AI and learning
quizforge
Turns any document into enriched Anki-style flashcards with a two-pass LLM pipeline. Built because reading twice is inefficient and forgetting is worse.
Audits data by day. Builds tools by night. Learns Kannada somewhere in between.
Data analytics for Internal Audit at Allstate India. GenAI audits, chatbots, SOX automation, event streams.
My job is asking systems to explain themselves. This one only needs a hover.
GenAI models, event streams and pipelines all orbit the same Sun: the data. Some come round daily, some once in years. An audit is knowing which is which, how far out each one runs, and what happens when one drifts. Hover a planet for its real year.
A career path best described as "plot twists, professionally documented."
Today I lead data management and analytics for Internal Audit at Allstate India. In practice that is three things: auditing GenAI use cases and designing the frameworks for it, building chatbots, and automating SOX in Microsoft Fabric. Around them sit data quality assessments for event-driven systems, advice to leadership on data strategy, and evaluations of ML pipelines from both a delivery and an assurance lens.
How I got here: I started out running a bank branch in Bihar, detoured through an MBA at IIT Roorkee, and spent years in Big 4 consulting building fraud analytics and audit dashboards, until I landed where data, risk and AI collide. Along the way I've mentored 20+ consultants, most of whom still talk to me.
Leading Internal Audit's data management and analytics practice: GenAI audits and audit frameworks, chatbots, event streaming assurance, SOX automation in Microsoft Fabric.
Embedded analytics into internal audit for global IT and e-commerce clients. Fraud detection in R and Alteryx, channel partner risk dashboards, revenue leakage hunts.
e-Governance projects for the Government of Karnataka.
End-to-end analytics delivery for international clients, from problem framing to dashboards.
Ran a full bank branch, topped the zone in car loans, and plugged revenue leakages the old-fashioned way: by reading everything.
Where the finance brain met the engineering brain. They've been arguing productively ever since.
Side projects, where "scope creep" is a lifestyle, not a finding.
AI and learning
Turns any document into enriched Anki-style flashcards with a two-pass LLM pipeline. Built because reading twice is inefficient and forgetting is worse.
Speech and desktop
A desktop transcription app with two-pass Hindi-English transcription on faster-whisper. Meetings end; transcripts are forever.
Markets and Python
A rule-based Nifty 50 swing-trade screener: Williams %R pullbacks, a 200-SMA trend filter, ATR-sized stops. The rules don't panic. I occasionally do.
Chrome and productivity
A Chrome extension for saving, searching and exporting your best LLM prompts locally. Because your best prompt shouldn't die in a chat you can't find.
The next one is still being scoped.
Longer thoughts on data, audit, AI, and whatever survives a second draft.
Money-specific writing lives at charpaisa.com.
Photographic evidence, admissible in casual conversation.
Disclosed voluntarily, in the interest of full transparency.
Chief Chaos Officer. Reports daily, escalates hourly.
Also second-guessing field placements from the sofa.
Cultivating patience, one wuxia chapter at a time.
The destination is negotiable. The playlist is not.
Progressive overload, tracked like everything else in my life.
I read supplement labels the way I read audit evidence.
A long-running engagement. Findings pending.
Last updated when this page was built. Auditor's honesty.
That concludes the engagement.
If the record reads cleanly, countersign it. Say hello.
Built in Bengaluru. No dashboards were harmed in the making of this website. © 2026 Abhishek Sinha. All findings final.
The record needs JavaScript. Nothing about this visit was kept.
Awaiting sign-off