All work

04 / 08AI System2026 – PresentFounder · open research startup

IndiEye

Market Intelligence Platform

Institutional-grade market sentiment for retail investors: news, social chatter and filings, read by a fine-tuned LLM and lined up against price action.

ML service News service
Status
Ongoing · 2026 – Present
Role
Founder
Backend
FastAPI · PostgreSQL · AWS
Model
Llama 3 (fine-tuned)

01 Problem

Institutional desks read the news, the chatter and the filings before they read the chart. Retail investors mostly don’t get that layer. IndiEye is an attempt to build it — in the open.

02 Approach

  1. Ingest the noise

    A real-time pipeline aggregates financial news, social media chatter from X and Reddit, and daily SEC filings.

  2. Read it with a domain model

    Llama 3, fine-tuned for domain-specific sentiment extraction, turns unstructured text into sentiment signals.

  3. Line it up with price

    Sentiment signals are correlated with price action using time-series models.

  4. Serve the insight

    A predictive dashboard in Next.js, backed by FastAPI on AWS with PostgreSQL storage.

  5. Split into services

    The ML layer (regime classification, strategy recommendation, key levels, backtesting) and the news & LLM-briefing layer run as independent, open-source microservices.

03 Results

Source streams
3
Fine-tuned for sentiment
Llama 3
Open-source services
2
  • Founded and open-sourced as an ongoing research startup; actively developed.
  • ML service exposes regime, strategy, key-level, market-radar and backtest endpoints.
  • News service aggregates multiple providers and RSS feeds into LLM morning briefs, priority feeds and sector briefs.

04 Stack

  • Next.js
  • FastAPI
  • PostgreSQL
  • Llama 3
  • AWS
Next case study · 05DataScout