🏦 Sell-Side Research Engine

A production-grade equity research pipeline that automates the full sell-side analyst workflow — from live data ingestion to institutional-quality research notes.

Mirrors the output standards of Goldman Sachs Tactical Research Group, JPMorgan Equity Research, and Morgan Stanley Research. Built for equity researchers, PMs, and quants who need institutional output at scale.

This project is open to contributions. Whether you want to add a new data source, improve valuation logic, or build the LLM reasoning layer — see CONTRIBUTING.md to get started.

What It Does

Given a watchlist of tickers and a date, the engine:

Fetches live fundamentals from SEC EDGAR XBRL, prices from Yahoo Finance, macro data from ECB SDMX + US Treasury feeds
Screens stocks across valuation, growth, quality, and momentum factors
Values top candidates using a multi-scenario DCF (bear/base/bull) + comparable company analysis (CCA) + reverse DCF to back-solve implied growth
Analyzes earnings call transcripts for management tone, guidance signals, and risk flags via LLM-driven commentary extraction
Quantifies risk — VaR, CVaR, Sharpe/Sortino ratios, beta, max drawdown, correlation matrices
Renders a 12-section GS-style HTML research note with embedded PNG charts, exported as HTML + PDF + Markdown

Key Features

📊 Research Automation

Multi-factor stock screener — valuation, growth, quality, momentum + analyst consensus overlay
Automated DCF valuation with CAPM-based WACC, sensitivity matrices (WACC × terminal growth), and football field charts
Reverse DCF — back-solves the implied revenue/earnings growth rate priced into the current market price
Peer comps (CCA) with sector-relative multiples and forward earnings
Catalyst calendar from market data and public disclosures
Optional discretionary analyst overlays via CSV (thesis, conviction, catalysts, risks, client angle)

🎙️ Management Commentary (NEW in v1.3.0)

Ingests earnings call transcripts (plain text or structured format)
Extracts structured signals: guidance tone, key themes, forward-looking statements, risk flags
LLM-driven summarization integrates directly into the research note narrative section
Supports pluggable transcript sources — paste raw text or point to a file path

🔄 Reverse DCF (NEW in v1.3.0)

Given current market price, solves for the implied FCF growth rate the market is pricing in
Outputs implied growth vs. analyst consensus to flag stretched/discounted valuations
Sensitivity table: implied growth across a range of WACC and terminal growth assumptions
Integrates with the existing DCF football field chart for a unified valuation view

🔌 Data Integration (No Synthetic Data)

Source	Coverage
SEC EDGAR XBRL	US fundamentals (revenue, net income, equity) — audited
Yahoo Finance	Prices, EPS estimates, analyst recommendations, price targets
ECB SDMX API	European interest rates, EUR/USD FX
US Macro	2Y/10Y/30Y Treasuries, VIX, DXY, WTI crude, gold futures

📁 Professional Outputs

HTML Research Note — 12-section dossier with GS navy/gold branding, embedded charts, scenario analysis
PDF Export — WeasyPrint rendering, print-optimized for client distribution
Markdown Daily Note — Git-friendly summary with top picks, macro snapshot, key changes
PNG Charts (150 DPI) — DCF sensitivity heatmaps, football field, factor radars, correlation matrices, sector heatmaps

⚙️ CLI Enhancements (NEW in v1.3.0)

--dry-run flag — validates all inputs and data sources without making live API calls
--quiet mode — suppresses verbose output for CI/automated pipelines
tqdm progress bars on all fetch and report-build stages for visibility into long-running jobs
PDF fallback rendering when WeasyPrint is unavailable

Quick Start

git clone https://github.com/DogInfantry/sellside-research-engine.git
cd sellside-research-engine
pip install -r requirements.txt

# Set SEC user agent (API etiquette)
export SEC_USER_AGENT="Your Name your_email@example.com"

# Fetch data + build full research note in one command
python main_v2.py build-all --as-of 2026-04-12

# Validate inputs without live API calls
python main_v2.py build-all --as-of 2026-04-12 --dry-run

# Run quietly for CI/automated environments
python main_v2.py build-all --as-of 2026-04-12 --quiet

Outputs to outputs/:

research_note_2026-04-12.html
research_note_2026-04-12.pdf
daily_note_2026-04-12.md

Windows users: Run from Anaconda Prompt (conda activate prime_quant) to avoid matplotlib DLL conflicts with Git Bash.

Architecture

sellside-research-engine/
├── main_v2.py                          # CLI: fetch-all | build-report | build-all | --dry-run | --quiet
├── trg_workbench/
│   ├── sources/
│   │   ├── sec_client.py              # SEC EDGAR XBRL parser
│   │   ├── ecb_client.py              # ECB SDMX API client
│   │   ├── macro_us.py                # US macro snapshot
│   │   ├── yfinance_client.py         # Yahoo Finance adapter
│   │   └── transcript_client.py       # Earnings transcript ingestion (NEW)
│   ├── analytics/
│   │   ├── screening.py               # Multi-factor composite ranking
│   │   ├── valuation.py               # DCF, WACC, peer comps, football field, reverse DCF (NEW)
│   │   ├── risk.py                    # VaR, CVaR, Sharpe, beta, drawdown
│   │   ├── catalyst.py                # Catalyst calendar builder
│   │   └── management_commentary.py   # LLM transcript extraction + tone scoring (NEW)
│   ├── reporting/
│   │   ├── charts.py                  # 12 PNG chart generators (GS palette, 150 DPI)
│   │   ├── pdf_renderer.py            # Jinja2 → HTML → PDF
│   │   └── templates/
│   │       └── research_note.html.j2  # 12-section GS-style HTML template
│   └── pipeline_v2.py                 # Full pipeline orchestration
├── data/
│   ├── analyst_views_template.csv     # Discretionary analyst overlay template
│   └── cache/                         # Cached fundamentals, prices, macro
└── outputs/                           # Generated research notes, PDFs, charts

Screening Model

Factor	Signals	Weight
Valuation	P/S, P/E — percentile ranked	30%
Growth	Trailing + forward revenue/EPS growth	25%
Quality	Net margin, ROE, Debt/Equity	25%
Momentum	1M, 3M, YTD price returns	20%

A second-pass Forward View layers in analyst consensus, price target upside, and earnings revision trends. Optional CSV overlays add discretionary conviction scoring.

DCF & Reverse DCF Valuation

Standard DCF inputs:

Free cash flow from SEC EDGAR XBRL
WACC = Risk-free rate (10Y UST) + β × 5% market risk premium − tax shield
3 scenarios: Bear (conservative growth), Base (consensus), Bull (upside case)
Sensitivity matrix: WACC × Terminal Growth Rate
Output: Intrinsic value per share + implied upside/downside to current price

Reverse DCF (NEW in v1.3.0):

Inputs: Current market price, base WACC, terminal growth assumption
Solves: Implied FCF growth rate priced in by the market
Output: Implied growth rate + comparison to analyst consensus; sensitivity table across WACC/terminal growth combos

Management Commentary (NEW in v1.3.0)

The management commentary module processes earnings call transcripts and extracts structured intelligence:

Guidance tone — positive / neutral / cautious classification
Key themes — revenue drivers, margin commentary, capex signals
Forward-looking statements — extracted verbatim with LLM tagging
Risk flags — supply chain, macro sensitivity, competitive threats

Output integrates into the research note narrative and is also available as a standalone JSON artifact.

Risk Metrics

VaR (95%, 1D) · CVaR · Beta · Volatility (21D/63D) · Sharpe Ratio · Sortino Ratio · Max Drawdown · Spearman Correlation Matrix

Customization

Change watchlist — edit WATCHLIST in trg_workbench/pipeline_v2.py

Adjust factor weights — edit WEIGHTS in trg_workbench/analytics/screening.py

Override DCF assumptions — edit valuation.py (rfr, market_risk_premium, terminal_growth_rate, tax_rate)

Add analyst views — copy data/analyst_views_template.csv to data/analyst_views.csv and populate the discretionary overlay fields:

Column	Type	Valid Values / Example	Description
`ticker`	string	`AAPL`	Stock ticker, matched case-insensitively against the research universe.
`stance`	string	`Buy`, `Hold`, `Sell`, `Overweight`, `Underweight`, `Positive`, `Negative`, `Neutral`	Analyst rating used in discretionary scoring.
`conviction`	number	`1` to `5` (`5` = highest conviction)	Analyst conviction score; the screening model normalizes this by dividing by 5.
`thesis`	string	`Services mix supports margin expansion`	Core investment thesis.
`catalyst`	string	`June WWDC AI updates`	Near-term upside or event catalyst.
`risk`	string	`China demand weakness`	Key downside risk or debate.
`client_angle`	string	`High-quality mega-cap defensiveness with AI optionality`	Client-specific framing for the callout section.
`management_access_note`	string	`Investor meetings requested after earnings`	Management access or meeting context for research notes.

Example:

ticker,stance,conviction,thesis,catalyst,risk,client_angle,management_access_note
AAPL,Buy,4,"Services mix supports margin expansion","June WWDC AI updates","China demand weakness","High-quality mega-cap defensiveness with AI optionality","Investor meetings requested after earnings"

Testing

pytest

Covers: SEC XBRL parsing · ECB normalization · screening logic · template rendering · risk analytics · chart generation · management commentary extraction · reverse DCF solver

🤝 Contributing

Contributions are welcome from equity researchers, quants, data engineers, and Python developers. Please read CONTRIBUTING.md for setup instructions, PR guidelines, and how to find good first issues.

For questions or ideas, open a Discussion or comment on an issue.

Roadmap

Items marked 🟢 are open issues ready to be picked up. See the Issues tab for full specs and acceptance criteria.

🏁 Milestone: v1.4 — Depth & Credibility

Make the research note defensible. These close the largest credibility gaps between a sophisticated demo and an institutional-grade engine.

Fundamentals & Filings

Valuation Engine

Output & Auditability

🏗️ Milestone: v1.5 — Infrastructure & Packaging

Make the repo look built, not hacked. These don't add features — they make every existing feature credible to a technical reviewer.

🔭 Backlog (no issue yet)

Analytics

Forward multiples in CCA (NTM EV/EBITDA, forward P/E)
LBO model stub — entry/exit with sponsor IRR
Event study module — abnormal returns around earnings and macro catalysts
Ranking explainability layer — factor attribution and sensitivity for screener output

Data Sources

OpenBB Platform SDK as optional aggregated source layer
SEDAR+ / Companies House for Canadian and UK filings
News catalyst detection — map headlines to catalyst calendar

Reporting

PowerPoint export (python-pptx) matching GS/JPM slide deck format
Excel DCF model export (openpyxl) with live formula links
Streamlit multi-ticker comparison dashboard

Engineering

Async data fetching (asyncio + aiohttp) to parallelise source calls
Redis-backed caching layer with TTL invalidation
Pluggable LLM backend (OpenAI / local Ollama / Mistral)

Want to tackle a backlog item? Open an issue to discuss scope before building.

Changelog

v1.3.0 — April 12, 2026

feat: Reverse DCF analytics — back-solves implied growth from current market price with sensitivity table
feat: Management commentary MVP — LLM-driven earnings transcript ingestion, tone scoring, and guidance extraction
feat: --dry-run CLI flag for input validation without live API calls
feat: --quiet mode and tqdm progress bars across all pipeline stages
feat: PDF fallback rendering when WeasyPrint is unavailable
docs: Analyst views schema documented

v1.2.0 — April 11, 2026

CLI refactor with quiet-aware progress bars and automated chart integration
Session output pipeline: research note, daily brief, session README
Analyst views schema documentation

v1.1.0 — April 11, 2026

--dry-run and progress bar features merged
.gitignore updated to exclude outputs/

License

MIT — free to use, modify, and distribute. See LICENSE.

Acknowledgments

Data: SEC EDGAR · Yahoo Finance · ECB SDMX · yfinance
Inspiration: Goldman Sachs TRG · JPMorgan Equity Research · Morgan Stanley Research
Stack: Python · matplotlib · seaborn · WeasyPrint · Jinja2 · pandas · yfinance

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

🏦 Sell-Side Research Engine

What It Does

Key Features

📊 Research Automation

🎙️ Management Commentary (NEW in v1.3.0)

🔄 Reverse DCF (NEW in v1.3.0)

🔌 Data Integration (No Synthetic Data)

📁 Professional Outputs

⚙️ CLI Enhancements (NEW in v1.3.0)

Quick Start

Architecture

Screening Model

DCF & Reverse DCF Valuation

Management Commentary (NEW in v1.3.0)

Risk Metrics

Customization

Testing

🤝 Contributing

Roadmap

🏁 Milestone: v1.4 — Depth & Credibility

Fundamentals & Filings

Valuation Engine

Output & Auditability

🏗️ Milestone: v1.5 — Infrastructure & Packaging

🔭 Backlog (no issue yet)

Analytics

Data Sources

Reporting

Engineering

Changelog

v1.3.0 — April 12, 2026

v1.2.0 — April 11, 2026

v1.1.0 — April 11, 2026

License

Acknowledgments

FilesExpand file tree

README.md

Latest commit

History

README.md

File metadata and controls

🏦 Sell-Side Research Engine

What It Does

Key Features

📊 Research Automation

🎙️ Management Commentary (NEW in v1.3.0)

🔄 Reverse DCF (NEW in v1.3.0)

🔌 Data Integration (No Synthetic Data)

📁 Professional Outputs

⚙️ CLI Enhancements (NEW in v1.3.0)

Quick Start

Architecture

Screening Model

DCF & Reverse DCF Valuation

Management Commentary (NEW in v1.3.0)

Risk Metrics

Customization

Testing

🤝 Contributing

Roadmap

🏁 Milestone: v1.4 — Depth & Credibility

Fundamentals & Filings

Valuation Engine

Output & Auditability

🏗️ Milestone: v1.5 — Infrastructure & Packaging

🔭 Backlog (no issue yet)

Analytics

Data Sources

Reporting

Engineering

Changelog

v1.3.0 — April 12, 2026

v1.2.0 — April 11, 2026

v1.1.0 — April 11, 2026

License

Acknowledgments