Franco Caccia

Franco Caccia Annual letter · 2026 Paper trading only

Franco Caccia: Quant. Software. Pilot.

Dear reader, this is a short letter about my work: a portfolio-managing bot that is still being measured, the projects I have shipped, and the record behind them. The bot trades on paper only, and every number about it here is a percentage.

Quant and software developer. M.S. student in Aerospace & Space Systems Engineering. Instrument-rated commercial pilot. Open to quant, software and aviation-tech roles.

Chapter 01

The bot

Live AutoTradingBot · Live, paper trading · January 2026 to present

Autonomous long-only portfolio manager in Go. Indicators, news and an LLM decide buy, sell or hold across stocks, ETFs, bond ETFs and cash. Bracket orders on Alpaca, risk limits enforced in code. Paper-tested with autonomous, real-time trade execution. Benchmarked against SPY and a 60/40 portfolio.

  • Go
  • Python
  • Gemini
  • Alpaca
  • Docker
  • Render

Limits live in code, not in the prompt.

Risk control 7 of 7. The others are in the notes below.

How it has done so far

Return
[+x.x%]
placeholder until the feed answers
vs SPY
[+x.x%]
placeholder until the feed answers
vs 60/40
[+x.x%]
placeholder until the feed answers
Max drawdown
[-x.x%]
placeholder until the feed answers

Placeholders. The bot's public stats fill them in when they load.

Policy bands, the share of the portfolio each may hold

  • Equities 40–75%
  • Bonds 15–40%
  • Cash 5–30%
Figure 1 · The bot, day by day, against the S&P 500 · Indexed to 100

Illustrative data, not real results. The bot's history feed is not published yet, so this chart shows a made-up series that starts at 100.

  • Bot, daily candle (filled = up day, hollow = down day)
  • S&P 500 (SPY)
  • 60/40 (SPY/AGG)
Portfolio vs benchmarks, indexed to 10030 trading days, Day 1 to Day 30. Last close: portfolio 101.67, SPY 98.18, 60/40 99.19. Start is 100. Illustrative data, not real results.97.5100.0102.5105.0107.5Day 1Day 30Day 16ILLUSTRATIVE DATABot 101.7SPY 98.260/40 99.2Bot minus SPY, by dayBot close minus SPY close for each of 30 trading days, in index points. Above the line the bot is ahead. Last day: +3.49. Illustrative data, not real results.+5.0−5.00Bot vs SPY
Portfolio vs benchmarks, indexed to 10030 trading days, Day 1 to Day 30. Last close: portfolio 101.67, SPY 98.18, 60/40 99.19. Start is 100. Illustrative data, not real results.97.5100.0102.5105.0107.5Day 1Day 30Day 16ILLUSTRATIVE DATABot 101.7SPY 98.260/40 99.2Bot minus SPY, by dayBot close minus SPY close for each of 30 trading days, in index points. Above the line the bot is ahead. Last day: +3.49. Illustrative data, not real results.+5.0−5.00Bot vs SPY
Portfolio vs benchmarks, indexed to 10030 trading days, Day 1 to Day 30. Last close: portfolio 101.67, SPY 98.18, 60/40 99.19. Start is 100. Illustrative data, not real results.97.5100.0102.5105.0107.5Day 1Day 30Day 16ILLUSTRATIVE DATABot 101.7SPY 98.260/40 99.2Bot minus SPY, by dayBot close minus SPY close for each of 30 trading days, in index points. Above the line the bot is ahead. Last day: +3.49. Illustrative data, not real results.+5.0−5.00Bot vs SPY

Illustrative. Day 30. The bot opened at 101.8 and closed at 101.7 (−0.18%), 3.5 points ahead of the S&P 500 at 98.2.

Open
101.85
High
102.09
Low
101.27
Close
101.67
Day change
−0.18%
SPY close
98.18
Bot minus SPY
+3.49

Index points, 100 on the first day. Click or touch the chart, use the arrows, or press the left and right arrow keys.

Table view
Illustrative series, not real results. Daily values indexed to 100 at the start. Percentages only.
DayOpenHighLowCloseSPY60/40Bot minus SPY
Day 1100.00101.6199.98101.38101.97101.13−0.59
Day 2101.38102.29101.07102.10101.91101.11+0.19
Day 3102.10102.94101.81102.63100.86100.52+1.77
Day 4102.63103.13102.30102.50100.56100.35+1.94
Day 5102.50103.84102.20103.69101.20100.73+2.49
Day 6103.69104.95103.39104.62102.16101.28+2.46
Day 7104.62104.97103.69103.74101.97101.18+1.77
Day 8103.74104.53103.60104.33101.14100.72+3.19
Day 9104.33104.52103.06103.10101.75101.07+1.35
Day 10103.10103.63102.96103.46102.50101.50+0.96
Day 11103.46103.86103.25103.73101.99101.22+1.74
Day 12103.73104.52103.46103.51101.50100.95+2.01
Day 13103.51103.72102.97103.38102.22101.37+1.16
Day 14103.38104.50103.11103.98102.49101.53+1.49
Day 15103.98104.00103.58103.82101.44100.94+2.38
Day 16103.82104.20103.47103.58101.42100.94+2.16
Day 17103.58103.58101.42101.74100.24100.28+1.50
Day 18101.74104.65101.70104.26101.60101.06+2.66
Day 19104.26104.49103.35103.40100.54100.46+2.86
Day 20103.40103.47102.59102.64100.12100.23+2.52
Day 21102.64103.07102.31102.91100.14100.25+2.77
Day 22102.91103.25102.63102.8999.4399.85+3.46
Day 23102.89103.44102.74103.1798.4799.31+4.70
Day 24103.17103.29102.54102.7198.2399.17+4.48
Day 25102.71102.99102.51102.7698.6999.45+4.07
Day 26102.76102.78102.26102.4198.6799.44+3.74
Day 27102.41102.49101.94102.0799.88100.15+2.19
Day 28102.07102.22101.62101.7199.5299.95+2.19
Day 29101.71101.90101.59101.8598.7699.52+3.09
Day 30101.85102.09101.27101.6798.1899.19+3.49
Figure 2 · Anatomy of a decision · An example, not a real decision
{
  "action": "BUY",
  "conviction": 7,
  "thesis": "Example only. A one-line reason goes here.",
  "reasoning": "Example only. The model cites the indicators and news it was given.",
  "suggested_tickers": []
}
  1. Model Replies with JSON like the example: an action, a conviction from 1 to 10, a thesis and its reasoning.
  2. Risk checks Code sizes the position from conviction, applies position and sector caps, and requires reward-to-risk of at least 2.
  3. Execution A market entry on Alpaca with server-side stop-loss and take-profit legs.

Notes: what the bot may not do

  1. Long-only: no shorting, margin, leverage or options.
  2. Position size scales with the model's conviction, under position and sector caps.
  3. Every entry is a bracket order: a server-side stop-loss and take-profit set from ATR.
  4. Reward-to-risk of at least 2 to 1 before an order is placed.
  5. Circuit breakers in the risk layer.
  6. Universe filters on exchange, price and volume. No leveraged ETFs or warrants.

Chapter 02

Other projects

  1. 01

    Shipped

    January 2026 to April 2026

    Stock-Quant-Reports

    Equity report generator for a student-managed investment fund: ticker in, two-page PDF out with weekly and monthly indicators, charts, narrative and a rule-based signal. Saves 5+ hours a week of manual analysis and feeds the fund's decisions.

    • Go
    • Python
    • React
    • MongoDB
    • Firebase
  2. 02

    Shipped

    October 2026

    DeFi lending protocol

    Over-collateralized lending protocol in Solidity using Foundry and Chainlink price feeds, with utilization-based interest, health-factor liquidations and explicit bad debt. Verified against an independent C++ model; fuzz, invariant and fork tests at 100% coverage. Educational project, not audited, never used with real funds.

    • Solidity
    • Foundry
    • Chainlink
    • C++
  3. 03

    Planned

    Planned

    MEV sandwich detector and mini blockchain

    A MEV sandwich detector and a mini blockchain, both in Go.

    • Go

Chapter 03

The record

Experience

  • August 2026 to present

    Finance & Data Research Assistant, Lehigh University, Bethlehem, Pennsylvania

    Build Python pipelines parsing N-CSR filings from SEC EDGAR into research-ready datasets. Extract securities lending revenue across two fund complexes and two decades of filing formats.

  • March 2025 to August 2026

    Quant Analyst & CS Intern, Student-managed investment fund, Lynchburg, Virginia

    Produced quantitative research on analyst-requested tickers for a ~$1.2M fund that outperformed the S&P 500 in 2025 at 0.6 beta. Built the Go/Python analysis system; co-managed the quant team's indicator-driven portfolio.

  • February 2024 to August 2026

    International Admissions Assistant, University of Lynchburg, Lynchburg, Virginia

    Evaluated and processed documents and guided incoming students with government documents.

Education

  • August 2026 to present

    M.S. Aerospace & Space Systems Engineering, Lehigh University, Bethlehem, Pennsylvania

  • January 2024 to August 2026

    B.S. Computer Science, minor in Mathematics, University of Lynchburg, Lynchburg, Virginia

Relevant coursework: Artificial Intelligence, Machine Learning, Data Structures, Databases, Operating Systems, Architecture & Assembly, Applied Regression, Accounting, Business Statistics, Investment Fundamentals, Finance.

Languages

Spanish (native), English (bilingual), Portuguese (fluent), Mandarin (basic)

Tools, by how often they were used in something built

Languages

  • Python 4
  • Go 3
  • C++ 1
  • Solidity 1

Frontend

  • React 1

Data

  • Firebase 1
  • MongoDB 1

Infrastructure

  • Docker 1
  • Render 1

APIs and models

  • Alpaca 1
  • Chainlink 1
  • Gemini 1

Tooling

  • Foundry 1

Everything on the resume: Assembly, C, C++, C#, Go, Python, Rust, Java, Node.js, React, Solidity, Foundry, TypeScript, PHP, MATLAB, MySQL, MongoDB, Azure, Docker, Google Cloud Run, Firebase, Linux, Git, Flex & Bison.