Autonomous Multi-Vertical Trading on Kalshi & Robinhood
By Colter Mahlum, Founder & CEO of Mahlum Innovations — published 2026-03-15.
Colter Mahlum architected and shipped this case study end-to-end. Mechanical engineer turned AI builder, he has personally led 11+ production AI deployments across manufacturing, healthcare, finance, and consumer apps from Bigfork, Montana.
24/7 Autonomous Trading Agent Across Sports, Weather, Crypto & Macro Markets
App: Axiom Trading Agent | Industry: Quantitative Trading | Client: Independent quantitative retail trader (event-contract markets) | Company size: Solo operator, ~$8.4K live bankroll | Duration: Multi-month iterative R&D, daily production iteration
Summary
An autonomous multi-vertical trading agent that finds, validates, and executes positive-EV trades on Kalshi prediction markets and Robinhood crypto end-to-end, with no human in the loop.
The Challenge
Prediction markets like Kalshi expose hundreds of new contracts daily across wildly different domains (a Yankees game, NYC's high temp, BTC's 15-minute close, the next CPI print). No single human can monitor every market, model fair value, size positions correctly, and execute fast enough to capture edge before it decays.
Our Approach
- Built per-vertical custom ML ensembles: LightGBM + LSTM + Transformer stack for crypto binaries, Elo + per-team ensembles for sports, calibrated regression for weather highs, macro models for CPI/GDP
- Layered a 6-persona GPT-4 swarm debate to argue each candidate trade and produce consensus probability + agreement score
- Added a final-veto LLM that re-evaluates the swarm-approved pick before any order is sent
- Implemented per-segment Platt/isotonic calibration with automatic ECE-based sizing-scale halts (the model self-throttles when predictions drift from reality)
- Built custom risk engine: live bankroll-aware Kelly sizing, per-vertical exposure caps, CLV tracking, duplicate-position dedup
- Deployed self-healing infrastructure: in-process supervisor + external watchdog that respawns the backend within 30s of any crash
Outcomes & ROI
- ~$8.4K — Live Bankroll: Managed fully unattended across ~27 concurrent positions in 6+ verticals
- 1,000+ — Signals Per Cycle: Narrowed from 1,000–1,200 positive-EV signals to ~15 candidates to 0–3 executed trades
- ~2 min — Full Cycle Time: Complete fetch → model → swarm → LLM → risk → execute loop runs every ~2 minutes, 24/7
Reduced operator intervention from 'restart backend whenever it crashes' to effectively zero. Hard 10%-of-bankroll exposure cap enforced at the API layer prevents any single bug from blowing up the account.
Technologies Used
GPT-4 Swarm, LightGBM, LSTM, Transformers, Platt/Isotonic Calibration, Kalshi REST + WebSocket, Robinhood Crypto API, AccuWeather API, Outlier.bet + Sportradar feeds
Key Takeaways
- Stacking specialized per-vertical models with an LLM debate layer outperforms either approach alone for noisy, multi-domain prediction problems
- Calibration drift detection (ECE-based throttling) is the unsung hero of unattended trading — it saves capital faster than any single trade decision
- Self-healing infrastructure isn't optional for 24/7 autonomous systems; a watchdog process is the cheapest insurance you can buy
Frequently Asked Questions
How does the agent decide which trades to take?
Each cycle pulls every open contract, prices it with a domain-specific model, runs a 6-persona LLM debate, applies a final-veto LLM, then passes the survivor through a Kelly-sized risk gate. Typically 1,000+ signals collapse to 0–3 actual orders.
What happens if a model starts drifting?
Per-segment calibration tracking automatically reduces position sizes (or halts a vertical entirely) when expected calibration error exceeds thresholds. The system self-throttles before bad predictions become losses.
Is it fully autonomous?
Yes — it runs end-to-end with no human in the loop. The watchdog supervisor respawns the backend within 30 seconds of any crash, and a 10% bankroll exposure cap is enforced at the API layer as a final safety net.
Related Services
About the Builder
Mechanical engineer turned AI builder. Colter personally architected and shipped Axiom Trading Agent end-to-end — strategy, model development, and the production MLOps work — alongside 10+ other AI systems across manufacturing, wealth management, healthcare, and consumer apps. Read full bio · LinkedIn.
More Case Studies
- Axiom AI — A multi-agent AI orchestration platform that lets users deploy, coordinate, and monitor swarms of specialized Claude/GPT/Gemini agents from a single command center to autonomously execute complex software, research, and operational workflows.
- Tier IV Pro — A unified manufacturing operations and project-management platform combining real-time bay scheduling, financial goal planning, quality/IDQ tracking, and AI-assisted forecasting into a single executive cockpit.
- Elysian Operations Platform (Time Warp + Nexus Wellness) — All-in-one workforce, payroll, wellness, and tax-compliance platform that replaces 8+ separate SaaS tools with a single unified system.
- Elysian Institute Website — A modern, SEO-optimized marketing website for a recovery, wellness, and longevity center, featuring 20 detailed service pages, transparent membership pricing, and direct integration with the client's external booking system.
- March Madness Commissioner — A full-stack commissioner dashboard that automates ESPN bracket league management — syncing picks, tracking payments, and generating AI-powered win projections — with a companion iOS mobile app.
- MealPlan — AI-powered family meal planning app that generates personalized weekly meal plans, tracks nutrition, scans barcodes, identifies food from photos, and acts as a personal nutrition coach.