MV ALTAIR · 142,000 DWTholmgaard line
move your cursor, she rocks
akaash
founding engineer at Kearney's AI Center of Excellence
product × engineering × ai
I build the AI platforms Fortune 500 procurement teams run on, and playful corners of the web that chase your cursor.
Some of these I built on my own; the rest are the AI tools I built at Kearney, the firm Fortune 500 companies pay to spend their billions more wisely. Same job either way: take a messy, expensive problem and ship the system that solves it.
01 / 10
Every registrationleaves a trail.
When you buy property in India, you trust brokers and listing sites that get paid to sell you something. This turns scattered government records into one honest picture, so a buyer, lender, or investor can decide on facts instead of a sales pitch. Built solo, end to end: the data pipeline, the maps, the logins, and six tailored views.
tn-property · live terminallive
rising districts
Coimbatore▲ +12%
Chennai S▲ +9%
Madurai▲ +6%
operational building planned6 workviews
02 / 10
Don't interview six.Interview the whole org.
Understanding how a big company's buying function really works used to take two or three weeks: a team running structured interviews, typing the answers into spreadsheet after spreadsheet, then analysts turning all of it into a diagnostic deck, and it only ever reached five or six people. This is a mobile-friendly web app you talk to out loud, so the same diagnosis scales across an entire procurement organization. It reasons alongside you as you speak, drawing on a built-in knowledge base to ask sharper follow-ups, working out how capable your team is and how it actually does things, mapping every answer to the specific hypotheses and factors it is testing, then pulling all of its interviews into one complete, org-level picture. Built solo, end to end: the realtime voice loop, the analyst brain underneath it, and the intake that routes every person to their own interview.
live discovery · how the org workslive
you · speaking
factors scored
13 / 20
probing next: how do approvals actually run?
ops lead“intake works fine”
the buyer“it's broken daily”
240 voices → one diagnosis, attributed
03 / 10
One engine.Every builder getsa premium site.
A property developer's only options for a website are a generic listings page that looks like everyone else's, or a pricey agency build that's stale the day flats start selling. This is an engine that spins up a premium, phone-first site for any builder in minutes, plus a dashboard to run their leads and live inventory. Built solo, on the same property-data brain as TN Property.
the collection · editorial luxury
SR Heights
Hosur · premium apartments
2 · 3 BHKRERA
fromEMI ₹38k/mo
Book a site visit
S.R & Co · CRM
24
leads
1.2k
views
template · skin · modules → one engine
04 / 10
Buyers ask the AI.Are you inthe answer?
More and more buyers don't open Google anymore. They ask ChatGPT or Gemini 'what's the best X for me?' and trust the answer. Most brands have no idea whether the AI even mentions them, what it says, or where it got that from. This measures exactly that across five AI assistants, finds where you're invisible and why, and turns it into a clear plan to fix it. Built solo, end to end: the data capture, the scoring, and the app.
perplexitychatgptgeminiai overviewsclaude
liveask · best wireless earbuds under ₹5k?
the AI names
01NorthPeak
cited · 2
02Vantage
cited · 1
03Olympia
mentioned
your brandnot named ✗
lever · third-party validation+ template ready
05 / 10
Forty spellings.One supplier.
A giant company records the same supplier 40 different ways across decades, so it cannot even answer 'how much do we actually spend with them?'. This AI tool reads millions of messy records and collapses 40 versions into one clean truth, so the company can finally negotiate from strength and cut waste. On one client it surfaced around $1.4B in savings.
ACME GMBH & CO. KG
Acme Gmbh
ACME GERMANY LTD.
a c m e gmbh…
ACME GMBH
golden record · 1 of 1.5M
06 / 10
Weeks of cleanup.One click.
Every consulting project starts with weeks of cleanup before real analysis can begin. I built the AI engines that do that cleanup automatically (sorting billions of rows of spending into categories, untangling messy supplier records) and the self-serve platform that put them in every consultant's hands at one click. Work that used to bottleneck on a few specialists now runs for the whole firm on demand.
PO_dump_final_v7.xlsx
spend_q3 (1).csv
vendors_RAW.xlsx
run
clean spend cube · 2h 14m
07 / 10
Provethe price.
A supplier says 'prices up 8%, blame steel and fuel' and most companies just pay, with no way to check. This breaks a product down into what it's actually made of and tests each piece against real market prices, so you can see exactly how much of a hike is justified and how much is padding. Built and owned end to end.
should-cost · spend → market~$401M
the cost mix sets expected · the market moves itthe gap · ~$19.3M
08 / 10
Read the market.Then the room.
A plan for how a company buys a whole category used to be only as good as whether a 20-year veteran was in the room. This captures how the best strategists think and lets anyone run that expert analysis, turning the market and the client's real strengths into a tailored, evidence-backed game plan. Built and owned end to end.
external category intelligence
Beroe · FRED · market indices
client maturity
scored across 29 themes
lever 01ready
lever 02ready
lever 03gated · build first
every call traceable to evidence
09 / 10
A day's research.Ten minutes.
Every buying decision starts with three questions: what should this cost, who else could we buy from, and what have similar companies saved? Answering took an analyst a full day of digging, and the first AI attempt just made things up. This answers all three in minutes, backed by real data from hundreds of companies, not guesswork. I led the build, including scrapping a first version that failed.
peer evidence · logisticsxlsx
what peers did about it
proposed
awarded
executed
~$234M realized ✓
1 day → 10 minevery answer cites its evidence
10 / 10
Read your contracts.Redraft them better.
A quick proof of concept for what AI can do with a company's whole pile of contracts, in one tool. It reads and extracts the key terms from every contract, sorts them into archetypes, and suggests a strategy for each. Then it redrafts: it checks your existing terms against a library of best-in-class clauses and proposes stronger, realistic ones to swap in, while a light pipeline tracks the updated contracts you send back to suppliers. Built solo in a couple of days, as a proof of concept, not a production system.
contract intelligenceproof of concept
MSA-2014archetype · framework
cl. 12 · payment termsp.7
net 30, no cap on increases
best-in-class
net 60, cap at CPI + 2%, audit right
extractsegmentstrategyredrafttrack
The web shouldplay back.
I grew up loving sites that answered back. So I build some. These are real, deployed, and waiting for your cursor. Go on, poke them.
MV ALTAIR · 142,000 DWTgo play
129knots
the phone leans your way
an.
an.
yes, it's watching your cursor too