Amit Kumar Pandey
Other versions
Email me

Amit Kumar Pandey

अमित कुमार पाण्डेय

CTO, Jellyfish TechnologiesProduct builderGame makerPhD scholar

Start the rideor scroll. Every stop is one part of my work.

I have been building software since 2008. I co-founded Jellyfish Technologies in 2011 and still run its technology. Around it I keep opening new lines: an astrology app, a trading platform, a games studio, and a PhD on where Delhi NCR should put its EV chargers.

My lines Jellyfish Technologies DashaGuru Anadi Algo PhD, Amity IIM Lucknow Tumblix Studios KonarkPro Sinki.AI

My work as a transit map Each line starts in the year that part of my work began: Jellyfish Technologies in 2011, Jellyfish Telecom in 2016, which became DashaGuru in 2025, KonarkPro in 2021, Anadi Algo in 2025, Sinki.AI and Tumblix Studios in 2026. IIM Lucknow is a short separate line from 2024 to 2025, now complete. The PhD at Amity University is its own line from 2026. Every line except IIM Lucknow still runs and meets where I am now.

Map not to scale. Select a line to jump to its station.

Namaste, I'm Amitनमस्ते, मैं अमित हूँ

Amit Kumar Pandey in a dark suit and tie, standing on the IIM Lucknow campus in front of a red-brick building IIM Lucknow campus

Where it started

शुरुआत

2008 to 2011

Engineer, then first-time founderNoida

Before 2008

MCA

A Master of Computer Applications, and the decision to write software for a living.

Before 2008

Learning Java

At a training institute, where I met Gaurav Chauhan. We were given work to do, and the two of us were the ones who always got it done.

  1. Senior Software Engineer, Damyant Software

    My first job, in Noida. Thirteen months of writing production code.

  2. Co-founded SAP Integrated Security & IT Services

    My first company. I ran it for two years, until a bigger idea came along.

Change here for Jellyfish Technologies

Gaurav and I start our own company. The first project pays $30.

Jellyfish Technologies

जेलीफ़िश टेक्नोलॉजीज़

Since 2011

Co-founder and CTONoida, Salt Lake City, Toronto, Sydney

It started with a $30 project.

Two Java students

Gaurav and I met at a training institute where we were both learning Java. We were given work to do there, and he and I were the two who always got it done. That is how we became a team.

A $30 project

We both took jobs after that. In 2011 we started our own company. The first project came through Elance and oDesk, the freelance sites of that time. It paid $30.

Twenty people on one account

After 2011 the bigger projects came: jBilling, Payleadr and many more. The jBilling work ran for two or three years, and at one point twenty of our people were on that one account. For two founders who had started from scratch, that kind of billing felt like heaven.

The month it stopped

Then jBilling was acquired. Work that had run for two or three years wound down within a month. We went from 70 people to 20. We were scared, and we had no idea what came next.

The dream we did anyway

Riding to Leh-Ladakh had been on our list for years, one of the things we had promised ourselves we would do in this life. The mess at work did not change that. We bought two motorcycles on loan, and in July 2016 we rode to Leh-Ladakh.

July 2016, in two photos

Amit and Gaurav in riding jackets beside a motorcycle on a mountain road, green valley and snow peaks behind
Gaurav and I on the road to Leh-Ladakh, July 2016.
Amit and Gaurav standing behind two new motorcycles outside the office
The two bikes, bought on loan, outside the office.

From 20 to 150

We came back, and new projects came too. We rebuilt from 20 people to 150, and what began with $30 has grown to $3 million a year.

  • $3Ma year in revenue, from a first project of $30
  • 150+people on the team today
  • 4offices: Noida, Salt Lake City, Toronto and Sydney
  • 5.0on Clutch, from 28 verified client reviews

What we build

Jellyfish builds software and AI for mid-market companies in the US, Canada and Australia, most of all in insurance and financial services. As CTO I own the architecture, the stack choices and our security posture.

I still build hands-on, like our document pipeline: OCR with EasyOCR and AWS Textract, an LLM to pull out entities, then GraphRAG on Neo4j to connect them.

  1. Founded in Noida

    Custom software for clients abroad, from day one.

  2. 2012 onwards

    The first big accounts

    jBilling, where twenty of our people worked on one account for two or three years, and Payleadr, an Australian direct-debit fintech we kept building for until 2021.

  3. Patra Corporation

    A US insurance services firm. We connected them to several top carrier APIs within two months, and they are still with us. On PatraPay, task time fell 85% and cost fell 75%.

  1. Heffernan Insurance Brokers

    APIs, middleware and single sign-on for a broker with 500 to 1,000 staff, and still with us eight years on. Their Clutch review cites a 95% on-time rate.

  2. Toronto office opens

    Our first office in Canada.

  3. AI and data modernization

    We modernize the systems and data our clients already run, then put AI into production on top of them, under SOC 2 Type II and ISO 27001.

Company facts

Founded
October 2011, Noida
Team
150+ people
Clients
120+ invoiced since 2018
Offices
Noida, Salt Lake City, Toronto, Sydney
Reviews
Clutch 5.0 from 28 verified reviews
Security
SOC 2 Type II, ISO 27001
Partner
AWS Select Tier
Stack
Node, Python, Go, .NET, Java on AWS, GCP and Azure

Some of the companies we build for: Patra Corporation, Heffernan Insurance Brokers, Payleadr and FM Integrated, among others. For Payleadr, an Australian direct-debit fintech, we shipped a mobile-first web app on a PCI-compliant payment gateway in three months.

jellyfishtechnologies.com

Two more lines from the same house

KonarkPro website

Since November 2021, Toronto

KonarkPro

Time tracking for teams

A desktop app for macOS, Windows and Linux that tracks tasks, apps and shifts, takes screenshots at random intervals, pauses when you go idle and syncs when you are back online. A web dashboard turns it into reports and invoices.

Electron, Preact and TypeScript; NestJS

konarkpro.com

Sinki.AI website

Co-founder since January 2026

Sinki.AI

Data and AI on Databricks

The Databricks practice of Jellyfish Technologies, launched in October 2025. It takes companies from raw data to data foundations to AI they can use.

sinki.ai

DashaGuru

दशा गुरु

Since 2025

Co-founderFormerly Jellyfish Telecom

Your very own Pandit ji, always by your side.

DashaGuru home page: Your very own Pandit ji, always by your side, with a phone showing daily guidance

From circuit boards to kundalis

Jellyfish Telecommunication began in March 2016 as a services company for IoT, firmware and PCB design. In 2025 it took a new direction and became DashaGuru, a Vedic Jyotish app for India.

You enter your birth details once. It works out your kundali from real planetary positions, and Pandit ji answers your questions about marriage, career, money and family, privately and without fear-selling.

How it works

  • Planet positions are calculated with the Swiss Ephemeris from your birth date, time and place.
  • A rule engine works out houses, dashas, doshas and timing windows. The words come after the maths.
  • Answers follow the classical texts: Brihat Parashara Hora Shastra, Phaladeepika, Lal Kitab and the KP Readers.
  • Guidance, not guarantees. Timing is shown as probability windows, and remedies are always optional.

dashaguru.comWeb app in betaAndroid in testing

Kundali screen in the DashaGuru web app, shown with demo data

Janm Kundali

Your birth chart

Twelve houses, divisional charts from D1 to D60, Vimshottari dasha, transits, and dosha and yoga checks. Shown here with demo data.

A sample page from the DashaGuru kundali report: your birth chart, the map you were born with

Kundali report

About 200 pages

Written as the story of your life rather than a data dump, in English and Hindi, as a PDF.

DashaGuru web app sign-in screen in Hindi

Ask Pandit ji

Around the clock, in 24 languages

Answers tied to your own chart and the classical rules. Pandit ji remembers what you asked before.

More inside

Daily guidance
Colour, food, direction and mantra for the day, plus a weekly planner.
Panchang and Muhurat
Tithi, nakshatra, choghadiya and Rahu Kaal for your city, and a muhurat finder.
Kundali Milan
36-guna matching, with Mangal dosha checks and cancellations.
KP astrology
The 249 sub-lord system for timing questions on marriage, career and children.

Also inside: remedies, including Lal Kitab remedies, and ratna advice that suggests a gemstone only when the chart supports it. Coming next: puja booking with a live webcast, and Pandit on call.

The line so far

  1. Jellyfish Telecommunication founded

    A services company for IoT, firmware and PCB design work.

  2. DashaGuru begins

    A new direction for the company: Vedic Jyotish, done with real calculations and respect for the shastras.

  3. Web app open in beta

    dashaguru.com and the web app are open to early users. The Android app is in testing.

Anadi Algo

अनादि एल्गो

Since 2025

Trading tools for Indian retail tradersInvite-only

Satta nahi, system.

Anadi Algo home page: Write your rules. Test them on history. Run them with guardrails.

Not a gamble, a system

With Anadi Algo a retail trader writes trading rules in plain Hindi or English, tests them on five years of market history, and runs them in paper mode before any real money is involved.

  • Strategy builder. Describe entry, exit and risk rules in your own words and get a rule draft you can test.
  • Backtester and paper trading. Try the rules on historical candle data first, then on live prices with no money at stake.
  • Scanner and options tools. A stock and F&O scanner, option chain tools and TradingView webhook input.
  • Guardrails. Stop-loss, limits and a kill switch, with broker connections for Zerodha, Dhan, Upstox and Groww.

Apply for early access

Paper trading and backtesting today. Live orders are not open to the public. Markets carry risk. No return promises. Not investment advice, tips or buy/sell calls.

Other experiments on this line

  • A desktop trading terminal in Go, with option chain and trade monitor panels (2025).
  • A reinforcement learning research model on Bank Nifty options data.
  • A cTrader bot in C# built around risk guards: a daily circuit breaker, spread filter and news blackouts.

IIM Lucknow

आईआईएम लखनऊ

2024 to 2025

Senior Leadership ProgrammeOrganizational Leadership

Back to class

Thirteen years into running technology at Jellyfish, I went back to class. From October 2024 to December 2025 I studied organizational leadership in the Senior Leadership Programme at the Indian Institute of Management, Lucknow, alongside the day job.

IIM Lucknow
Indian Institute of Management, Lucknow
Senior Leadership Programme
Organizational Leadership
Oct 2024 to Dec 2025
Completed alongside full-time work

PhD, Amity University

पीएचडी, एमिटी यूनिवर्सिटी

2026 to 2029

Part-time PhD scholar, Information TechnologyAmity Institute of Information Technology, Gurugram

An Integrated Optimization Framework for EV Adoption and Public Charging Infrastructure in Delhi NCR

Delhi NCR is adding public EV chargers without a forecast of how many EVs each area will have or how much charging they will need. Adoption differs sharply by city and by vehicle type, so some chargers sit idle while busy areas get queues and overloaded transformers.

My research builds a data-driven method that forecasts adoption and charging demand for each part of NCR, then recommends where to put chargers and how many.

The method, in three steps

  1. AdoptionForecast new EVs for each RTO and vehicle type with Bass and Gompertz curves, under low, medium and high scenarios.
  2. DemandSpread the forecast over 1 km² hexagonal zones and turn it into daily kWh. Compare ARIMA, XGBoost and LSTM on open data.
  3. SitingScore candidate sites with GIS and AHP, choose them with p-median and a genetic algorithm, size each site with an Erlang C queue, then check transformer limits.
  • 6,85,398EVs registered in Delhi NCR from January 2021 to August 2026, about 1 in 11 new vehicles. This is from the public VAHAN extract used in my approved synopsis; the chart that follows uses my own later pull from all 56 NCR registration offices, so its totals run higher.
  • ~164EVs for every public charging station in Delhi, and that is a lower bound.
  • 88% vs 35%Share of new e-rickshaws and autos that were electric in 2025: Delhi against neighbouring Gurugram.
  • 160Charging stations OpenStreetMap shows across all of NCR. The central government (PIB, August 2025) lists 1,967 public stations in Delhi alone.

New EVs registered in Delhi NCR, per year

Bar chart. 2021: 37,489. 2022: 96,156. 2023: 1,31,382. 2024: 1,54,979. 2025: 1,84,181. 2026 up to 15 November: 1,86,280. 2021: 37,489 2022: 96,156 2023: 1,31,382 2024: 1,54,979 2025: 1,84,181 2026, to 15 November: 1,86,280 37,489 96,156 1,31,382 1,54,979 1,84,181 1,86,280 202120222023202420252026*
New EVs registered in Delhi NCR per year
202137,489
202296,156
20231,31,382
20241,54,979
20251,84,181
2026*1,86,280
*2026 counts run to 15 November. Source: VAHAN, all 56 NCR registration offices, pulled 15 Nov 2026.
Map of Delhi NCR with public charging sites from open data: dense in Delhi, almost none in Gurugram, Noida, Ghaziabad and Faridabad

What open data can see

Charging sites visible in open data: dense in Delhi, almost absent in the satellite cities.

The line so far

  1. Registered

    Batch 2026 to 2029, January session.

  2. H1 2026

    Coursework done

    Research methodology, and research and publication ethics.

  3. Literature review

    49 studies, plus government documents and data sources. A first pilot on open charging data from Shenzhen: gradient boosting cut forecast error by about 30% against repeating the same hour yesterday.

  4. Synopsis approved Latest

    About 30 minutes with the research committee, presentation and questions. The topic and synopsis are now final.

Next on this line

  1. Next

    First conference paper

    “Can S-Curve Models Forecast Electric Vehicle Registrations? A City-Level Study of Delhi NCR Using VAHAN Data”. Draft ready, now with my supervisor for review.

  2. 2027–28

    Build and test the framework

    Adoption, demand and siting models, checked against a uniform grid of chargers for 2030 and 2035.

  3. Aug 2029

    Thesis

    Earliest submission date for a part-time PhD.

Research details

Supervisor
Dr. Paras Chawla, Professor and Director, Amity School of Engineering & Technology
Co-supervisor
Dr. Amandeep Sharma, Senior AI/ML and Generative AI Lead, Accenture
Institute
Amity Institute of Information Technology, Amity University Haryana, Gurugram
Data
VAHAN registrations, Delhi's charging list, OpenStreetMap, UrbanEV and ACN-Data, DERC tariff orders

Tumblix Studios

टम्बलिक्स स्टूडियोज़

Since 2026

Co-founderGames for Roblox, iPhone and iPad

Heists, dinosaurs, superheroes and one very hungry mouth.

The studio

Tumblix Studios is the games studio I co-founded. Eight of our games are live on Roblox and one is on the App Store. We write the code, build the worlds, and tune every game on what real players do in their first minute.

Roblox Studio and LuauUnity 6Blender

Visit tumblixstudios.com

Shopping Cart Mayhem

Heist on Roblox

Grab a cart, rob the store, outrun security. One fix to the first escape lifted the share of new players who make it out from about 24% to 76%.

Play on Roblox

Dino Wrangler: Tame & Ride

iPhone and iPad

Steal eggs, hatch 19 dinosaurs and ride them. Plays offline, with no ads and no in-app purchases. Available in 175 countries and regions.

Get it on the App Store

Black Veil

Co-op zombie shooter on Roblox

A story shooter for one to four players, now three chapters long.

Play on Roblox

Pet Dealer

Animal sim on Roblox

Catch, ride and sell 104 animals, from a puppy to a T-Rex, in a world built from voxels.

Play on Roblox

Rift City Heroes

Superhero action on Roblox

Fly across a 2.5 km city with four powers and stop the Rift invasion in a three-chapter story.

Play on Roblox

+1 Mouth

Simulator on Roblox

Play as Munch, who eats and grows across 15 towns.

Play on Roblox

COMMAND: The Last Decision

War action on Roblox

Ten-minute squad battles through a ruined town, with scopes up to 8x thermal and a warship boss.

Play on Roblox

Moonbell Town

Co-op adventure on Roblox

Up to four friends restore a moonlit town's lost light across six rescue adventures.

Play on Roblox

Car Wash Tycoon

Tycoon on Roblox

Start with one bucket and a $1 wash. End with a neon street of businesses.

Play on Roblox

Products

प्रोडक्ट्स

built end to end

Working products I build alongside Jellyfish

Jellyfish is a services company: we build other people's software. These are the products I build alongside it, each one working end to end.

CloseTrace lead recovery dashboard: forms started, left without sending and sent

CloseTrace

Trace every visit. Close more leads.

  • Replays every visit to your site, with friction and errors flagged.
  • Counts every lead form a visitor started but never sent.
  • Ranks what is costing you leads: slow pages, failing forms, script errors.
  • Heatmaps for clicks, rage clicks, dead clicks and scroll depth.

TypeScript, NestJS, Next.js, PostgreSQL

closetrace.comLive

Velqiri home page with two iPhone screens showing recovery and a coach insight

Velqiri

Your health, connected. Your next step, clearer.

  • Turns Apple Watch sleep, heart-rate and workout data into recovery, sleep and strain estimates.
  • Compares today with your own last four weeks, not with other people.
  • A coach on every screen that says what changed and gives one next step.
  • A native Apple Watch app: day summary, one-tap start, guided breathing.

Go, Expo, native watchOS. A wellness app, not a medical device.

velqiri.comBeta

Kavach console: a case with its linked alert, response actions and their progress

Kavach

Cybersecurity for every laptop and server in the company.

  • Every enrolled device in one list: health, operating system, agent version, site and group.
  • USB, application and data-loss policies, rolled out in stages: 1%, then 10%, then everyone.
  • Alerts scored by severity and tagged with the matching MITRE ATT&CK technique.
  • Turn an alert into a case and respond from it: isolate the device or kill the process, with a second person approving risky actions.

Rust agent, Go, React, PostgreSQL, ClickHouse

Working demo

TenderAssist tender feed with fit, partial and not fit verdicts for each tender, shown with demo data

TenderAssist

Know which government tenders you can actually win.

  • A daily feed of tenders from GeM and state e-procurement portals, each one stamped fit, partial or not fit for your company.
  • A requirement-by-requirement check: what the tender asks, what you have, and the exact clause it comes from.
  • A document vault for certificates, financials and work orders, with expiry warnings.
  • Models only read the documents. A rule engine decides eligibility, so every verdict can be traced.

Go, PostgreSQL, React

Beta

5dollardevs project page with plan, build, review and ship tabs, shown with demo data

5dollardevs

Software shipped by AI developer agents, in the open.

  • Bring an RFP or answer a guided interview, and get a plan before anything runs.
  • Tickets with dependencies and an estimate in points, hours and cost. Nothing runs until you approve.
  • Pick a team from 1,000 named AI agent profiles across five lanes: design, build, peer review, verification and DevOps.
  • Agents open branches and pull requests on GitHub. A reviewer agent can block the merge, and a human clears the gate.

Go, React and TypeScript, Python, PostgreSQL

In development

Inference console playground with an answer from an open model and its speed figures

Inference gateway

One API in front of open models.

  • Works with the OpenAI and Anthropic SDKs by changing the base URL.
  • A playground that shows speed and token counts for every answer.
  • A request log with time to first token, tokens per second and cost for each call.
  • API keys per project, with limits and usage by day, model and key.

Go, Next.js, PostgreSQL, Redis

In development

Desktop transcription app showing a waveform and a Hindi transcript

Offline transcription

Speech to text that never leaves the computer.

  • English and 12 Indian languages, in native script or as an English translation.
  • Trims the recording and cleans up hum and background noise first.
  • An editable transcript, exported to Word, PDF, text or subtitles.
  • Runs fully offline on a desktop, with batch mode for many files.

Python, Whisper models, PySide6

Beta

More from the workshop

Spectrogram of a generated Hindi voice sample

Hindi calling voice

Voices for phone calls

Natural Hindi and English voices for phone calls, in calm, cheerful and apologetic tones, on an open voice model that runs locally. Listen to the same line in two tones:

नमस्ते, कॉल करने के लिए धन्यवाद। आपका ऑर्डर कल शाम तक पहुँच जाएगा।

Python, VoxCPM2

Prototype

Layer view of a generated memory chip layout

Chip layout tool

SRAM layout and design-rule checks

Generates a memory array layout, writes it as a real GDS file and checks it against a reference rule set.

C++, Python, KLayout

Prototype

Drone control dashboard with a simulated camera feed and telemetry

Drone autonomy

Scout drone, in simulation

Type or speak one command, Hinglish included, and a simulated scout drone flies a search pattern, streams its camera and files a mission report.

Python, ArduPilot, Gazebo

Prototype

Records of a research run: thirty reports and thirty cross-checks

Research orchestrator

Two agents, one checked answer

Two AI agents research the same question separately, review each other's reports, then merge everything into one document. One saved run: 15 tracks, 30 reports, 30 cross-checks.

Python

Prototype

Film still: Shiva and Parvati seated, child Ganesha standing, Mount Kailash behind

Ganesh aur Kartikeya

3D animated short in Hindi

The story of the fruit of wisdom, told on Kailash. A one-minute test is finished; the full film is in production.

Blender

In production

Events and notes

कार्यक्रम और नोट्स

Since 2022

Where I show up, and what I write down

Events

  1. PhD synopsis, Amity University Haryana

    About thirty minutes with the research committee, questions included. Approved.

  2. ConnexFM 2023 national conference

    Jellyfish showcased FM Integrated, the facility-management platform we built for a client in Maryland.

  3. Collision 2022, Toronto

    With my co-founder Gaurav Chauhan.

All my lines meet here.

संपर्क करें

If you are building something hard, in software, products, games or research, write to me.

amit@amittechguy.inLinkedIn

More adventure ahead.

अभी और एडवेंचर है

No one knows what comes next. But I am ready to take it all.