Live at nepsescanner.comFinTechData platformAI agent
NepseScanner: turning raw exchange data into trustworthy market intelligence.
An end-to-end market-data and analytics platform for the Nepal Stock Exchange, built for investors and active traders. It covers automated data collection, analytics, a web app and mobile PWA, alerts and an AI research assistant.
38Scheduled data pipelines running automatically
280Secured REST API endpoints
6,800+Automated tests (backend + frontend)
81Database tables with full data provenance
The challenge
Market data that was scattered, messy and hard to trust
Exchange prices, trade-by-trade records, broker activity, company reports and economic data came from many sources and in many formats, including PDFs and spreadsheets. Investors had no single reliable place to see what was happening, and most "signals" on the market were not tested against evidence.
Our solution
One verified data platform with analytics and AI on top
- Automated collectors that validate every record and log its source.
- A fast analytics layer powering live dashboards, scanners and broker-flow views.
- An AI research assistant that only reports numbers produced by real calculations.
- Alerts, a mobile app, subscriptions and admin tools, all in one system.
Feature explorer
Everything we built
Market in 30 seconds
The whole market, explained in plain words every day.
Intraday market chartThe NEPSE index through the session against yesterday's close, with money traded, trades and companies.
Busier than usualCompanies trading several times their normal volume, flagged automatically.
Rises vs fallsMarket breadth at a glance: how many stocks rose, fell or stayed flat.
Coming upBook-close dates, lock-in expiries and other important dates for the companies you follow.
Stock Intelligence
Filter every listed company using facts from the exchange's own trade records, not predictions.
Plain-language conditions"Easy to trade", "busier than usual", "above their average price", "haven't jumped too fast".
Explains what was excludedShows how many companies each condition removed.
Saved presetsCombine conditions over 5, 10 or 20 sessions and save them as a preset.
Caution flagsUnusual trading patterns are marked, with a clear note that a flag is not an accusation.
Rankings, minute by minute
Live leaderboards through the trading session and over the last 20 sessions.
Live intraday listsRising most, falling most, most money, most shares, busier than usual.
20-session viewsStrongest, pullbacks, volatility, drawdowns and unusual moves.
Fair by designThin trading can't game the lists: a company must trade at least Rs 1 million today to rank.
Filter by industryHydropower, banks, microfinance, hotels and more.
Every company, in plain words
A page for every listed company, written for ordinary investors.
Price history1-month, 3-month and 1-year charts with clear notes on adjustments.
"What kind of stock is this?"Trend, ease of trading and price steadiness compared with every other listed company.
What happened todayAn automatic plain-English summary of the day's move.
The evidenceEvery measurement, with what it can and can't tell you.
Test a strategy on real history
Choose a rule and see what your money would have become, after every fee and tax.
No-code backtestingPick a company, a period (back to 2016) and a rule such as "the price drops sharply".
Realistic costsResults include every fee and tax, and execution is realistic: next-day opening price.
Honest comparisonShown side by side with simply holding the company or keeping cash.
Liquidity guardOnly companies that trade enough can be tested.
Built for the phone in your pocket
An installable app (PWA) for retail investors, with built-in lessons.
Mobile-first appToday, Explore, Rankings, My stocks and company pages are all designed for phones.
Learn the basicsShort lessons on shares, dividends, bonus shares, charts and diversification.
Glossary built inMarket words explained wherever they appear.
Guest modeTry everything instantly without signing up.
The engine underneath
What runs behind the screens, every trading day.
38 automated data pipelinesPrices, order book, indices, trade-level records, corporate actions, fundamentals, and central-bank data parsed from PDFs and spreadsheets.
Provenance on every numberEach record stores its source, collection time and quality score. Original responses are kept so data can be reprocessed.
AI research assistantA tool-using LLM agent for internal research. It runs on a private on-premise model and blocks any number that no tool produced.
Self-healing operationsWatchdogs, failure alerts to Telegram, verified daily backups with restore tests, and release checks that fail loudly on any mismatch.
On mobile
Designed for the phone first
Architecture
How it fits together
Sources
Exchange feedsprices · depth · trades
Company reportsfundamentals · actions
Macro dataPDF · Excel · API
Automation
38 scheduled collectorsvalidate · normalise
Provenance & qualitysource, time and score on every row
Data
PostgreSQLoperational store
DuckDBanalytics warehouse
Feature storeversioned
Intelligence
AI Quant Desktool-using LLM agent
Research enginebacktests · regimes
FastAPI · SDK280 endpoints
Delivery
Web dashboards80+ screens
Mobile PWA
Telegram alerts
Technology
Tech stack
Backend & data
Python 3.12FastAPISQLAlchemy 2AlembicPostgreSQL 16DuckDBpandasNumPy
Frontend
React 19TypeScriptViteRechartsPWASEO pre-rendering
AI
LLM tool-use agentOllama (on-premise)Qwen 2.5Anthropic Claude
Automation & ops
Dockersystemd timersCloudflareGitHub ActionsTelegram Bot API
Quality
pytestVitestPlaywrightruffpip-audit
Integrations
eSewaKhaltiGoogle sign-inPDF/Excel parsing
What this means for your business
The same capabilities, applied to your data
Workflow automationThe same pipeline engine can collect your sales, inventory or finance data automatically, every day.
Live dashboardsThe same analytics stack can show your KPIs in real time, with alerts when something changes.
Trustworthy AI agentsThe same guarded-agent design can answer questions about your business without inventing facts.
Next project
View