BIST AI
An analytics system that processes financial data through validation layers.
Autonomous Data Flow & Dashboard Hub
Forget the static data in standard finance apps. A high-performance command center rendered with Next.js, powered by autonomous bots (Playwright) and job queues (BullMQ) that track the pulse of Borsa Istanbul.
Ruthless Portfolio Algorithms
An analytical judge that eliminates emotional and amateur decisions. Thanks to complex calculations running on PostgreSQL, it instantly analyzes your asset allocation, sector concentrations, and portfolio health, warning you against potential crashes.
Dynamic Synthetic Reports
It systematically scans metrics such as momentum, RSI, and moving averages to present analysis reports with visible target and risk levels.
Scalable SaaS Infrastructure
This is not just an interface; it's a robust subscription ecosystem based on Supabase and TypeScript. A full-fledged B2B/B2C infrastructure that seamlessly manages API limits, database loads, and background processing power for different user profiles.
Engineering Flex & Production Metrics
Empirical benchmark numbers achieved in high-concurrency production.
Tracked Equities
Real-Time Market Data Across All BIST Equities
API Latency
Redis In-Memory Hot Cache Layer
Technical Indicators
Vectorized Multi-Period Mathematical Engine
LLM Hallucination Rate
Deterministic Financial Guardrails & Strict JSON
Autonomous Balance Parsing
Automated Quarterly Financial Statement Parser
End-to-End System Architecture Flow
From raw target DOMs to sub-50ms reactive delivery.
Live Market & Disclosure Ingestion
Asynchronous stream ingestion pulling price feeds, market depth, and corporate disclosures across 500+ equities.
Python / FastAPI Analytics Engine
Vectorized mathematical processing computing RSI, MACD, Bollinger Bands, and moving averages in milliseconds.
Deterministic Ratios & LLM Analysis
Deterministic validation of all financial ratios (P/E, P/B, EBITDA) fed into structured JSON schemas for AI commentary.
PostgreSQL & Redis Hot-Cache
Historical time-series persisted in PostgreSQL; hot analytical snapshots cached in Redis for <20ms query velocity.
Reactive Next.js Financial Terminal
Dynamic Recharts charting, instant stock screening, and interactive financial scorecards delivered with zero UI lag.
Deep Engineering Case Studies
How architectural challenges were solved with algorithmic precision.
Sub-20ms API Response via Redis Hot Cache Across 500+ Equities
The Critical Challenge
Calculating real-time indicators directly from relational SQL queries during market hours caused 800ms+ latency spikes and severe CPU bottlenecks.
Implemented Solution
Engineered a tiered Redis TTL caching strategy. Pre-computed analytical snapshots are stored in memory, invalidating only upon verified new tick intervals.
Verified Engineering Impact
Slashed API response times from 800ms to under 20ms (97.5% improvement), handling concurrent requests effortlessly.
Vectorized Indicator Pipeline: 10+ Technical Indicators in Milliseconds
The Critical Challenge
Traditional Python iterative loops computing multi-period RSI, MACD, and Bollinger Bands took several seconds across large ticker universes.
Implemented Solution
Rebuilt the calculation pipeline utilizing vectorized array operations (NumPy/Pandas) integrated into non-blocking asynchronous FastAPI endpoints.
Verified Engineering Impact
Computed 10+ complex technical indicators across the entire market universe in single-digit milliseconds.
Zero-Hallucination AI: Deterministic Guardrails for Financial Statements
The Critical Challenge
Feeding raw balance sheets directly into generic LLMs frequently causes fabricated financial ratios, distorted EBITDA calculations, and dangerous hallucinations.
Implemented Solution
Constructed a strict two-stage pipeline: a deterministic mathematical engine first computes exact financial metrics, then feeds verified structured JSON payloads to the LLM solely for qualitative executive synthesis.
Verified Engineering Impact
Achieved a 0% hallucination rate in balance sheet scoring, delivering enterprise-grade financial summaries.
Autonomous Corporate Disclosure & Quarterly Statement Parser
The Critical Challenge
Public disclosure reports arrive with inconsistent accounting item naming, nested table layouts, and irregular publishing schedules.
Implemented Solution
Engineered an intelligent parser that reconciles disparate accounting line items, calculates year-over-year (YoY) and quarter-over-quarter (QoQ) deltas, and stores normalized models.
Verified Engineering Impact
Quarterly statements are ingested, parsed, and scored within minutes of publication with 100% autonomy.
Technologies Used
Related services
Explore the service areas demonstrated by this project.
Interested in this project? Let's work together.
Contact