> initializing profile.shashank
> reading visitor… verdict: good taste
> status OPEN — FOR THE RIGHT ONE
> rendering environment_
00
PRODUCTION-GRADE SYSTEMS
DATA ANALYST
AI ENGINEER
FASTAPI · PYTHON
BIG DATA · HPC
LLM AGENTS
Shashank
HI, I'M SHASHANK
Production
Grade Systems
Data
Analyst
AI
Engineer
// I DON'T SHIP HOPE. I SHIP SYSTEMS.

Anyone can ship a demo. I ship things that pass audits, take traffic, and don't call you at 3 a.m.

Data doesn't lie. People just don't ask it the right questions. I do — and I show my work.

Eight AI agents applying for jobs while their owner sleeps. Not a pitch deck — a priced, live product I built alone.

↓ SCROLL. I'LL DO THE TALKING.
000°ORBIT
Make the call
// THE READ
Let me guess why you're here. You're a…
// READ 01

You've read forty profiles today. They all say "passionate" and "results-driven." Here's what they don't say: I've supported validated systems where a mistake ends up in front of a regulator, and I've built three products alone that real people use. You don't need to guess whether I can do the work. You need to decide how soon you want it done.

// OR PROVE YOU CAN READ A ROOM
Shashank headshot
OPEN — FOR THE RIGHT ONE2026
// THE FILE ON ME

Shashank. Remember the name.

Regulated pharma IT taught me that "it works on my machine" is not a sentence you get to say. Founding three companies taught me nobody is coming to do it for you. Put those together and you get an engineer who ships production-grade systems, reads the data before making the call, and builds AI that does real work — and who would rather show you than tell you.

Audit-proof21 CFR PART 11 · GXP
Ships soloTHREE LIVE PRODUCTS
Reads the dataMSC BIG DATA & HPC
// THE ARSENAL

Tools I don't have to Google.

Nothing here is something I've only read about. Every one of these has been used on something that's live.

// HOW I OPERATE

Four rules. No exceptions.

// ROOT 01

Ship like a regulator is watching

Because in pharma, one was. Versioned, validated, monitored — those habits don't switch off when the company gets smaller.

FastAPI & SQL
// ROOT 02

Read the room, then the data

Opinions are cheap. I build the pipeline, look at the numbers, and then make the call — in that order, every time.

Python & Pandas
// ROOT 03

Automate the boring, own the important

If a machine can do it at 3 a.m., a machine should. My agents apply, reach out, and report back. I handle judgment.

Agents & LLMs
// ROOT 04

Build it yourself first

Three companies, zero co-founders. I know what every part of the machine costs because I've paid for each one.

Product & GTM
// FLAGSHIP VENTURE

Novrant AI — most people apply. I built the thing that applies.

An autonomous job-application system I founded and built alone. It finds the roles, tailors and submits the applications, and reaches the humans behind the postings — while the candidate sleeps. I've been on the other side of the table. I built the fix.

STARTERPROELITE
Visit novrantai.com →
0AI AGENTS
0+JOB BOARDS + LINKEDIN
24/7AUTONOMOUS RUNS
// 01  MINDSET

Job hunting is a systems problem, not a motivation problem.

Telling people to "apply more" ignores the math. The bottleneck is throughput and follow-through — exactly the kind of problem you solve with engineering, not willpower. I built the system I wished existed when I was on the other side of the table.

// 02  THE BIGGER PROBLEM

Hiring is automated on one side only.

Employers screen with ATS filters, keyword matching, and AI. Candidates still copy-paste the same details into hundreds of forms by hand, get ghosted, and never reach a real person. Good people lose not because they're unqualified but because the process is stacked against them.

// 03  THE SOLUTION

Give candidates the same automation advantage employers already have.

Novrant AI runs eight specialised agents in a pipeline: discovery across LinkedIn and 20+ boards, fit scoring against the candidate's profile, application tailoring, form submission through browser automation, and direct outreach to hiring contacts with MX-validated email addresses — so messages land in real inboxes, not the void.

DISCOVERSCORETAILORAPPLYOUTREACHTRACK
  • Built with Python, Playwright, FastAPI and LLM agents — no off-the-shelf wrapper.
  • Every application is logged with a full audit trail; regulated-systems habits applied to a consumer product.
  • Internal ops platform for monitoring runs, quality and customer accounts.
  • Three tiers — Starter, Pro, Elite — priced by volume and outreach depth.
// 04  VISION
A world where every qualified person gets seen — where opportunity depends on what you can do, not on how many forms you can fill.
// 05  MISSION
To put a tireless, intelligent agent on the side of every job seeker, so the effort of getting hired matches the effort it took to become worth hiring.
// THE TELL — A GAME ON THIS PAGE

Everyone has a tell. Spot mine.

The whole site is the board. I show you a section, the screen blinks, and one thing has quietly changed. Find it. Three lives, ranks from Paralegal to Name Partner. Nobody has played this on a portfolio before — and it tells me exactly how well you read a room.

WORKS ON PHONE · NO SIGN-UP · BEST SCORE SAVED
// EXHIBITS

Evidence, not adjectives.

// PROJECT 01 AI Automation

Novrant AI

Autonomous job-application service. Eight coordinated AI agents apply across LinkedIn and 20+ job boards, with MX-validated email outreach to HR contacts. Tiered SaaS with an internal ops dashboard.

PythonPlaywrightFastAPILLM Agents
LIVE →
// PROJECT 02 Web Product

Winter Arc

Discipline-challenge web app with registration, daily habit tracking, a public leaderboard, and an admin analytics dashboard. Server-rendered with HTMX for speed, deployed on Railway.

FastAPISQLiteHTMXRailway
CODE →
// PROJECT 03 Regulated Systems

MODA at AstraZeneca

Application support for a validated lab data system under 21 CFR Part 11, ALCOA+, and cGMP. Change control, data integrity, and incident handling in a GxP environment.

GxP21 CFR Part 11SAPInfra
ENTERPRISE
// PROJECT 04 Consumer Brand

Gram & Graft

Indian clean-protein brand built on honesty and zero fillers. Landing page, domain setup, and launch — designed and shipped end-to-end.

ReactVercelDNS
LIVE →
// PROJECT 05 Personal AI

J.A.R.V.I.S.

Modular personal assistant: boot agent, mail intelligence, plugin registry, and a live dashboard. Architecture-first build for a system that runs the day.

PythonAgentsDashboard
IN PROGRESS
// NEXT

Always looking at the next thing.

Three companies in, still building like it's day one. The best work I've done is whatever I'm shipping next — and I'd like it to be for you.

// THE CLOSE

You've seen enough. Let's talk.

You read this far, which means one of two things: you have a role, or you have a problem. Either works. The form is short on purpose.

// payload_preview.json
sender:  [Awaiting Name]
email:   [Awaiting Email]
message: "[Awaiting Message]"
TRANSMITTED ✓
// GET READY ROUND 1 STREAK 0 SCORE 0 0.0
// THE TELL

Everyone has a tell.
Can you spot mine?

The whole site is the board. I'll show you a section, the screen will blink, and one thing on the page will have changed — a number, a word, two chips swapped, a label. Find it.

  1. Memorise the section while the timer runs.
  2. After the blink, tap the thing that changed.
  3. From round 4, sometimes nothing changes. Call it — "NOTHING CHANGED" — and be right.
  4. Three lives. Streaks multiply. Rounds get faster and subtler.
// VERDICT
RANK
Associate

0SCORE
0ROUNDS
0YOUR BEST
Make the call