Available for work

I build software that makes complicated business work simple.

From AI and automation to internal tools, dashboards, websites, and custom platforms, I build around the way your business actually works.

What I do

I build what your business actually needs.

Sometimes that means automating a messy workflow. Sometimes it means building a tool your team has been missing. Sometimes it means taking a pile of data and making it useful. I usually start with the problem, then figure out what needs to be built.

Automation & AI

Make the repetitive work disappear.

I build systems that connect your tools, handle the steps in between, make decisions, and keep work moving without someone having to do it all manually.

Websites & Internal Tools

Build the software your team actually needs.

From websites people use to internal tools your team relies on, I build software around the way the work actually happens instead of forcing the process into a generic template.

Data & Analytics

Make your data useful, not just available.

I build dashboards, reporting systems, data pipelines, and analytics tools that bring scattered information together and make it easier to see what's happening and what needs attention.

Selected work

Real projects. Real systems. Real results.

Each project comes with a short screen recording of the actual product in action.

  1. Project 1

    Custom platformAlgo trading

    One platform for every strategy

    Research markets, build strategies, backtest them, and trade live, across crypto, US, and Indian markets, with multiple brokers and accounts.

    What makes it interesting

    • Build & test strategies

      Create strategies, backtest them, and move from paper to live trading.

    • One platform, multiple markets

      Crypto, US, India, exchanges, brokers, and accounts in one place.

    • Real-time everything

      Positions, P&L, alerts, reporting, and market data as it happens.

    • Built for scale

      Handles thousands of instrument calculations on every tick without making the platform feel slow.

    • Analyse before you trade

      Technical and fundamental analysis built directly into the platform.

    Business result

    From researching an instrument to running and monitoring a live strategy, the entire trading workflow happens in one place.

    Tech stack

    • Next.js
    • React
    • TypeScript
    • Python
    • Go
    • FastAPI
    • PostgreSQL
    • DuckDB
    • Parquet
    • Apache Arrow
    • WebSockets
    • Broker APIs
    • Kite Connect
    • Binance
    • Telegram
    • Caddy
    • AWS

    Every screen · 10

    • Algo trading: Overview

      Overview

    • Algo trading: Strategies

      Strategies

    • Algo trading: Positions

      Positions

    • Algo trading: Market data

      Market data

    • Algo trading: Instruments

      Instruments

    • Algo trading: Brokers

      Brokers

    • Algo trading: Funds

      Funds

    • Algo trading: How a strategy is written

      How a strategy is written

    • Algo trading: Notifications

      Notifications

    • Algo trading: System

      System

  2. Project 2

    Automation · AI agentBrowser automation

    One request. A complete job search.

    A browser-based AI agent that searches multiple job platforms, finds relevant openings, and turns them into a structured dataset.

    Recorded on the real extension, the real portals, and the real dashboard.

    Before

    Searching each job platform separately, checking openings manually, and copying relevant jobs into a spreadsheet.

    After

    Ask once. The agent searches across platforms, evaluates the results, and returns the relevant openings in one structured spreadsheet.

    What makes it interesting

    • 321 roles in under a minute

      Role, company, location, posting date, and the original link for each one, ready to export.

    • Works across 34 job boards

      Ashby, Greenhouse, Lever, Wellfound, LinkedIn, and more. It picks the right approach for each one.

    • Learns new sites on its own

      When it encounters a site it hasn't seen before, it explores how it works and keeps notes for the next run.

    • Local or paid AI models

      Choose between local models for privacy and cost control, or paid models when you need more capable reasoning.

    Business result

    The team starts with a complete, structured list of relevant openings instead of manually searching across multiple platforms.

    Tech stack & architecture

    Architecture

    A Chrome extension with a side panel manages the search. Each platform runs in its own tab group, while the agent communicates with its server over WebSockets to handle the search and processing.

    Built with

    • Chrome extension
    • JavaScript
    • Node.js
    • WebSockets
    • Playwright
    • OpenAI
    • Local models
    • Public APIs
    • Database
    • Nginx
    • AWS
  3. Project 3

    AIRecruiting research tool

    One company. Everything you need to know.

    A recruiting research tool that turns a company's website into a complete view of its people, open roles, and recent activity.

    Recorded against the real tool, exactly as it shows a company.

    Before

    An hour per company, jumping between their website, LinkedIn, careers page, and news before anyone reaches out.

    After

    Paste their website and hit run. In about half a minute, one page shows who they are, who to contact, every open role, and the latest news.

    What makes it interesting

    • Who to contact, one click away

      Recruiters, HR leads, engineering heads, and founders, each linked straight to their LinkedIn.

    • Every open role in one list

      From their careers page and job boards. Filter to engineering, India, or roles posted this week.

    • News worth opening with

      Funding, launches, and expansion, so the first message has something relevant to say.

    Business result

    Outreach starts with the right people, the open roles, and a reason to get in touch already on one page, instead of an hour of research first.

    Tech stack & architecture

    Architecture

    A Node.js web app that pulls people from Apollo, company pages through TinyFish, and open roles and news from public job boards and feeds, all at the same time. An OpenAI model sorts through the information, and the result is saved so the next lookup is instant.

    Built with

    • HTML & CSS
    • JavaScript
    • Node.js
    • OpenAI
    • Public APIs
    • Database
    • Nginx
    • AWS
  4. Project 4

    WebsiteEvent experience

    One link. Everything attendees need.

    A lightweight event website built to give thousands of attendees the information they need without forms, accounts, or unnecessary steps.

    Speakers are blurred in the recording. Their names are theirs to share, not mine.

    Before

    Event details sat in posters, forwards, and group chats, so people kept asking when, where, how much, and how to get there.

    After

    One link has it all: date, time, fee, what to expect, and directions, with one tap to share it, save it to the calendar, or find the way.

    Business result

    Attendees get every answer from one link they can forward, without having to register or fill in a form just to get the information.

    Tech stack

    • HTML & CSS
    • JavaScript
    • Python
    • Caddy
    • Google Maps
    • AWS
  5. Project 5

    Trading toolBroker automation

    One press. Entry, stop-loss and target.

    A trading interface that works across major Indian brokers, combining order placement, position sizing, and automated trade management into a single workflow.

    Recorded against the project's own dev harness: the real extension, the real logic, a mock broker. Nothing in this clip places an order, because demoing it live would mean putting real money into the market.

    Before

    When the market moves fast, working out the quantity and then placing the entry, stop-loss, and target one by one costs seconds and invites mistakes.

    After

    Pick the contract and the capital to use. One click places the entry, stop-loss, and target together, with the quantity already worked out.

    What makes it interesting

    • Works across major Indian brokers

      Connect supported broker accounts and use the same workflow for order placement and trade management.

    • Sized for you

      Enter the capital to use and it works out the quantity from the lot size and live price.

    • Automatically protected

      Set the stop-loss as a percentage or an exact amount. If you don't set one, a predefined safety floor is applied automatically.

    • Locks in profit as it moves

      Once the trade goes your way, the stop-loss trails behind it automatically.

    • Big orders handled

      Large orders are split into chunks the exchange will accept.

    Business result

    The trade is sized, entered, and set up with its stop-loss and target in one workflow, reducing the manual steps needed when the market is moving fast.

    Tech stack & architecture

    Architecture

    A React and TypeScript side panel inside Chrome, built with Vite and Tailwind, backed by a FastAPI server that connects to supported Indian brokers through their APIs and keeps a trade journal in SQLite. It runs behind Nginx on AWS.

    Built with

    • Chrome extension
    • React
    • TypeScript
    • Vite
    • Tailwind CSS
    • Python
    • FastAPI
    • Broker APIs
    • WebSockets
    • SQLite
    • Nginx
    • AWS
  6. Project 6

    AutomationFinance workflow

    Payment in. Invoice out.

    An automated invoicing and payment-tracking workflow built around Google Workspace.

    Recorded on a rebuild with a made-up academy and made-up students, because the real sheet holds real people's names, emails, and payments.

    Before

    Every enrolment meant working out tax, balance, and the next due date, building the invoice, and emailing it by hand, then doing it all again at every instalment.

    After

    Type the name, course, and amount into the sheet and click Invoices. The student gets the invoice by email, and each new instalment updates the same invoice.

    What makes it interesting

    • The sheet does the maths

      Discount, tax, total, balance, and the next due date fill themselves in as you type.

    • One invoice that keeps up

      Each instalment updates the same invoice number with what's paid and what's left, and the PDF matches.

    • The right email every time

      A welcome email with the invoice on the first payment, the right message at every stage after, and reminders before each due date.

    • Daily visibility

      The team gets a daily Slack summary showing what's been paid and what's still pending.

    • Works inside the tools they already use

      It runs inside Google Sheets and uses the organisation's existing invoice design, with no separate system to manage.

    Business result

    Nobody types an invoice by hand. Each payment updates the right invoice and email, while the team can see what's paid and pending without checking everything manually.

    Tech stack

    • Google Sheets
    • Apps Script
    • JavaScript
    • Google Slides
    • Drive
    • Gmail
    • Slack

About

I like making complicated things feel simple.

I'm Salmon Paul. I build automations, AI systems, internal tools, and software for businesses that have too much manual work going on.

Portrait of Salmon Paul
Salmon Paul

I'm usually interested in the problems that sit somewhere between “we just do it this way” and “there has to be a better way.”

Maybe someone is copying the same information between three different tools every day. Maybe a team is spending hours researching companies. Maybe a process depends on one person remembering ten different steps.

Those are the things I like working on.

I figure out how the work actually happens, break it down, and then build the system around it. Sometimes that's a small automation. Sometimes it's an AI agent, an internal platform, a dashboard, or an entire workflow from scratch.

I started doing this for myself because I was constantly building little tools to make my own work easier. Eventually, I realised the same approach works really well for businesses too.

No agency. No handoffs. You work directly with me.

I handle the thinking, building, testing, and fixing myself. That also means I care about the boring parts that are easy to overlook: what happens when something fails, what happens six months later, and whether the person using it actually finds it easier.

The goal isn't to add AI or automation just because it's possible.

The goal is to make the work better.

  • Based in

    Hyderabad, working with anyone whose timezone I can survive

  • Usually a good fit

    Founders, small teams, and businesses with a process that feels unnecessarily complicated

Have something that takes too many steps?

Tell me how it works today. I'll help figure out what can be automated, rebuilt, or simply made less painful.

A 15-minute chat is usually enough to understand what you're trying to solve, talk through what's possible, and see if I can help.