// Built during my role as Senior AI/ML Engineer at Brainfog Technologies
IRIS India
I built a full-stack, multi-agent AI platform for NuqiWealth that acts like a team of specialized research analysts for the Indian stock market (NSE and BSE), covering fundamentals, technicals, news sentiment, and stock screening.
Tech Stack
How it works
Seven specialized agents split up the analysis, fundamentals, technicals, sentiment, ranking, screening, broad search, and portfolio, with built-in Shariah-compliant investing filters and index-based filtering.
- Built seven specialized agents: fundamental, technical, sentiment, ranking, screener, broad search, and portfolio, each owning a different slice of the analysis instead of one model doing everything.
- Added Shariah-compliant investing filters and index-based filtering (Nifty 50, Nifty 500, Sensex).
- Built a morning scheduler that automatically analyzes hundreds of stocks: running cross-agent analysis, storing results in the database, running sector-level analysis and a health checklist per sector, and backtesting the technical agent's own decisions against the last 5 years of data.
- Tuned signal thresholds based on real backtest results rather than assumptions.
Results
Specialized Agents
Exchanges Covered (NSE, BSE)
Faster After Backend Optimization
Before
A comprehensive cross-agent query, like "Is Reliance a good buy?", took up to 2 minutes. Every specialized agent wrote out a full analysis independently, an orchestrator combined those into one essay, and a formatting guardrail rewrote it again for brand consistency.
After
A 4-phase backend optimization collapsed that into a single high-quality synthesis pass at the end of the pipeline. The same comprehensive analysis now takes 49-55 seconds (about 55% faster), single-agent queries like fundamentals or technicals answer in 20-25 seconds, and general questions get a dedicated fast-path that skips the agent pipeline entirely, answering in 3-8 seconds. Sentiment analysis, previously the slowest single-agent query at up to 90 seconds, now answers in 35-45 seconds. Averaged across all query types, response time fell from roughly a minute to under 30 seconds.
Why I built it
Retail investors researching the Indian market usually have to piece together fundamentals, technicals, and news sentiment from different tools, with no easy way to filter for Shariah-compliant options or validate a signal before trusting it.
How I approached it
I split the research work across specialized agents instead of one generalist model, then built a dedicated backtesting engine so every signal is checked against real historical data before its thresholds are trusted.
Full System Architecture
This is the actual, code-verified architecture, not a simplified diagram. It covers the main orchestrator (26 nodes), all five specialized sub-agents, the portfolio integration, the infrastructure layer, and all four background scheduler jobs. Explore each part below, drag to pan and scroll or pinch to zoom.
