The horse's step
Vivek
Raghunathan
A horse crosses the board in jumps and can still land on every square exactly once. Nothing skipped. Nothing repeated.
I've spent about 20 years building systems held to rules like that. Now I'm building Adalma AI: fraud and financial-crime protection for credit unions and community banks.
Engineer · Founder · Angel investor
36 squares · 1 path · 0 repeats
01 Now
Building Adalma AI
Founder and CEO. Fraud and financial-crime protection for credit unions and community banks, with evidence an examiner can check.
“Fraud is personal to me. After about 20 years building systems where failure is not an option, I kept watching vulnerable people get taken by scammers. Credit unions are often the first to see it, so I built one platform a small team can run to catch it early.”
From adalma.ai
- Fraud and financial crime
Cernio
Gives the teller the hold and what to ask.
adalma.ai/cernio → - Collections and member voice
Cura
Stops collecting when a member names a scam cue.
adalma.ai/cura → - AI governance
Tessra
A kill switch on every model call.
adalma.ai/tessra →
- One audit trail
- Hash-chained. An examiner verifies it without us.
- Your cloud, your keys
- Single-tenant. Member data never leaves it.
- Your core
- Symitar and Fiserv DNA adapters, written to vendor spec.
- Oct 2026Sponsored the ABA AML & Fraud Conference, Arlington
- Oct 2026Accepted into NVIDIA Inception
- Oct 2026Joined the Cartesia Startups grant program
- Jul 2026Started building Adalma
02 The line so far
Two decades, one line.
Device drivers, education platforms, AI inside banks. The line draws as you read.
-
2008 to 2009 · Sunnyvale, CA
Juniper Networks
DVT Engineer
- Design verification and testing of switch control and interface boards
-
2009 to 2013 · Santa Clara, CA
Kno, Inc. (acquired by Intel)
Senior Software Engineer & Software Lead
- #2 non-founding employee; engineering grew to 80 through the Intel acquisition
- Embedded Linux, device drivers and kernel work on tablet hardware
-
2013 to 2016
Intel
Backend Engineering Manager & Tech Lead
- Led backend for Intel Education Study, serving 100K+ students
- A catalog of 1M+ titles and 100M+ data points
- Led an Intel Education Service Corps assignment in Rwanda; team excellence award from the Rwanda Education Ministry
-
2016 to 2018 · San Francisco Bay Area
Togg
Founding team, Head of Engineering
- Built engineering from the founding team out
-
2018 to 2026 · San Francisco Bay Area
Founding CTO
- Co-founded it and ran it for eight years alongside the Decision Minds CTO role: B2B SaaS for customer success, 25 customers at peak
-
2018 to Present · San Francisco Bay Area
CTO & Client Partner
- Eight years putting data and AI to work inside large financial institutions
- Replaced HRMS, CRM, recruiting and expense SaaS with in-house platforms on one data layer, saving close to $60K a year
- Keta runs the employee lifecycle: payroll, expenses, approvals, e-signatures and SharePoint
- Time and expenses from WhatsApp, with a Teams bot for the rest
- A CRM with an Outlook plugin, a LinkedIn Chrome extension and Apollo enrichment
-
2026 to Present · San Francisco Bay Area
Founder & CEO
- Fraud and financial-crime protection for credit unions and community banks
- Cernio takes a case from alert to a filing-ready SAR. Cura stops collecting at a scam cue. Tessra puts a kill switch on every model call
- Hash-chained audit trail, single-tenant in the institution's own cloud: adalma.ai/proof
- First customer in production since August 2026
- ABA AML & Fraud Conference sponsor, NVIDIA Inception, Cartesia Startups grant
B.E., Electronics and Communication, Anna University · M.S., Electrical Engineering, Florida International University
03 Builds
After hours.
Trading systems, agents and apps. Each gets a sigil generated from its name.
- Running
SIBT
Autonomous trading infrastructure on Cloudflare Workers: scheduled scans, approve-to-execute orders and daily briefs, run from a phone. Every strategy is preregistered with a bar to clear. In October, eight books were retired on their numbers and four kept.
dossier.sibt.ai → - Paper, under review
typesafe-daytrader
An intraday options agent on a paper book. Options flow picks the names, a model answers typed questions about each, and plain code enforces the risk box. Up 1.6% against SPY's 1.2% from September 17 to October 28. Review on November 17.
- Live
Chillax
An IPTV player for the living room, built for family. Live on the Amazon Appstore for Fire TV. The Android TV build ships from GitHub with Google Play on the way, and the Apple app is in TestFlight for iPhone, iPad and Apple TV.
Releases → - Shipped
Refusal Rails
Solo entry to the Alpaca x lablab.ai AI Trading Agents Hackathon, built around the orders it declines: 22 named risk gates and 564 logged refusals. The paper book finished four live sessions up 8.46%.
GitHub → - Open source
traderkit
Claude Code as a trading terminal: MCP tools for risk gates, signal ranking, fill checks and routing across SnapTrade and TradeStation. A three-model trading council and a 401-test suite.
GitHub → - Live
FareScout
A Chrome extension that decodes opaque travel deals: the hotel behind a Priceline Express Deal, the airline behind a mystery fare. Every guess is checked against Hotwire, ITA Matrix and Google.
Chrome Web Store → - Live
should-i-be-trading
One answer before the open: TRADE, CAUTION or NO TRADE. The Copilot tier proposes covered calls, puts them to a five-model vote, and places them through your own broker only after you approve.
sibt.ai → - Live
WorldMonitor
A geopolitical and macro dashboard, ported from Vercel to Cloudflare Workers. Its shipping, trade and economic signals feed the SIBT books every morning.
worldmonitor.sibt.ai → - Open source
snaptrade-trade-mcp
An MCP server for SnapTrade trading: equities and multi-leg options.
GitHub →
Retired on the numbers
Every experiment gets a scoring date and a bar to clear. These missed. The write-ups stay public: a lesson kept to yourself helps no one.
- Sep 11, 2026 AI Hedge Fund Seven persona agents debating each position. 47.6% across 698 picks against a 50% bar.
- Oct 2026 Autopilot, live sleeve A mechanical auto-trader with an AI council and drawdown halts. It lost to SPY. Its rails now guard the books that remain.
- Oct 28, 2026 Options wheel, paper Cash-secured puts and covered calls with 50% buy-backs. Down 0.6% while SPY rose 0.7%.
- Oct 28, 2026 Four more paper books A general book, a data-center power thesis, trend-day and overnight movers. None beat its benchmark.
- Aug 2026 banger-board A 13-signal pop screen. A 48% win rate, with losers two to three times the winners. Kept as a screen, retired as alpha.
04 Backed
Founders I've backed.
Angel investing since 2019, mostly infrastructure and AI. A member of The Chennai Angels since 2020, across the US and India.
13 companies · 2 ecosystems · 3 through The Chennai Angels
-
Ample Foods
Consumer / food
-
Automation Anywhere
RPA / automation
-
Cerebras Systems
AI compute
-
Cohesity
Data infrastructure
-
Ember Fund
Fintech
-
Rubrik
Data security
-
ThoughtSpot
Analytics
-
CricClubs
Sports tech
-
INSIDE.COM
Media
-
The Indus Valley
Consumer / D2C
-
Mangopoint
Agritech / supply chain
-
Trelish
Consumer
-
Unplex
Media tech
05 Perspective
On the AI wave
Every platform shift rewards the people who ship through it. Mobile and cloud did. This one differs in one way: trying an idea now costs almost nothing.
Agents are the new backend. Not chatbots. Pipelines of small, verifiable steps with gates and audit trails. That’s how I build trading systems, and how we build compliance systems at Adalma.
Regulated industries are the frontier. Anyone can demo AI. Making it hold up to an examiner, policy-gated, logged and tamper-evident, is where durable value sits. The institutions that need AI most can afford it least. Closing that gap is a business.
Taste survives automation. When anyone can generate anything, the scarce skill is knowing what’s worth building and when to say no.
The next decade
Satya Nadella sees SaaS thinning into agent-run workflows. Elon Musk keeps pulling the debate down to physics: intelligence per watt, compute as manufacturing. Nikesh Arora asks the question every enterprise will face: how do we find, govern and stop our agents? Put together: agents do the work, physical AI does the labor, and governance decides who gets to sell it to serious institutions.
Energy is the commodity. Every one of those futures runs on electrons, and on water for cooling. Watch where the power and water deals land; that’s where the compute goes.
Robots come home. The wave that matters most socially is domestic: machines that help seniors live on their own for another decade. I’ve spent years on the people fraud targets first. Dignified eldercare may be the most consequential thing this wave ships.
Agrotech is the one I’m most excited about. Autonomous robots that sweep fields at night with calibrated UV light instead of chemical pesticides: no residue, no runoff, nothing for pests to build resistance to. TRIC Robotics already runs them on California strawberry farms. It’s where AI, robotics, energy and water meet, and it gets almost no headlines.
06 Writing
Thinking out loud.
Recent posts on fraud, compliance and AI in regulated institutions.
- Oct 29, 2026 · LinkedIn
Six months into building Adalma, the surprise hasn't been the technology. It's been time. The people we build for, BSA officers and fraud teams at credit unions and community banks, are some of the busiest people I've met. Every…
Read the post → - Oct 27, 2026 · LinkedIn
Every AI-in-AML pitch I hear leads with the auto-close rate. 40% of alerts closed with no human. Sometimes more.
Read the post → - Oct 22, 2026 · LinkedIn
Two pieces of news for Adalma this week: we've been accepted into the NVIDIA Inception program, and we've joined the Cartesia Startups grant program. Adalma builds AI for institutions that carry a big bank's compliance load…
Read the post → - Oct 20, 2026 · LinkedIn
In an elder exploitation SAR, the account holder is usually the one person who doesn't belong in Part I. It's an easy mistake to make. The case opens on the member's account, so the member's name lands in the subject field.
Read the post →
From the hackathon build log on X
Repo is public. Autonomous options agent on a Cloudflare Workers cron, no laptop in the loop. REST owns orders, Alpaca's MCP server owns options data, and both answers get logged so the integration…
Sep 3, 2026 on X →Day 1 of my Alpaca hackathon agent trading autonomously in the cloud. It placed zero option orders. Four gates blocked it, each only visible after fixing the one before: 1. sleeve off via an…
Aug 31, 2026 on X →The pitch: trading agents get judged on the orders they place. Mine is built around the orders it declines. 22 named risk gates sit between every decision and every order, and every gate that fires…
Aug 28, 2026 on X →
07 Your turn
Your move.
Jump like a horse: two one way, one across. Visit every square once.
Tap any dot to share something. Each step, everyone who has it tells their neighbors.