Triple Berry Solutions

Products and lab

Built in house. Run in production.

Every product here is designed, engineered and operated by Triple Berry Solutions. They are how we learn what agentic AI can really do with money, markets and media, before we recommend it to a client.

Launching · Trading with Burhan

AlphaSqueeze Option Dashboard

An options-income intelligence platform that treats every premium sold as a reduction in cost basis, not a directional bet. Under the hood is a deterministic, citation-backed edge engine built from peer-reviewed quantitative research; on top is an AI coach, market intelligence and a strategy tracker.

  • Cost Basis Reduction Framework: adjusted cost per share is the primary number, not raw P&L
  • A 107-edge catalog with page-level citations, regime gates, and a 41-claim list we refuse to encode
  • Every recommendation shows what was considered and rejected, with the reason
  • Momentum Wheel strategy tracker, alerts, market brief and an AI coach on Claude
  • Two tracks: a workshop tool for retail traders and a Bloomberg-style institutional product for family offices and RIAs
Next.js · ReactClaude APIRule engineDexie · RechartsRunnable skills
In production since January 2026

Cash Flow, the AI CFO agent

A Claude Code agent that turns a folder of monthly bank statements into a fully reconciled cash flow workbook. Ten accounts across four banks, categorized, consolidated into an income statement and balance sheet, and validated by three checks that must equal zero before the month closes.

  • One command per month; closed months are never rewritten, so corrections survive
  • Categorization resolves through learned reference numbers, exact matches, rules, then a flagged fallback
  • Every manual correction is harvested on the next run, so each decision is made once
  • Sign conventions, transfers and liabilities handled the way an accountant expects
  • Extends naturally to small business bookkeeping, family offices and property portfolios
Claude CodePythonopenpyxlLearned rulesCSV · PDF · Numbers
In production · Burhan Kunda Visuals

Content Creation Engine

A skills-driven pipeline that turns raw footage or stills into finished, on-brand video. Six brand tokens govern every graphic. Claude runs the rough cut, graphics, AI b-roll, finishing and export, and a growing library of generated sound and music is reused across every job.

  • Two pipelines: a reel engine for stills and AI clips, a video editor for narrated footage
  • Taste lives in one place (brand and style files); jobs are disposable and never fork it
  • Generative video and image models for b-roll, product shots and title cards
  • Delivered work for restaurants, listings, travel and finance creators
Claude skillsFFmpegGen-AI videoBrand tokensSFX library

The lab

Research and working systems.

Smaller projects that test one idea each: an agent architecture, a market hypothesis, a connector. Some became products; some were honest failures that taught us where the edge is not.

Working system

Options Portfolio Monitor, four layers

Live data and Black-Scholes greeks, portfolio analytics with IV rank and expiry radar, a deterministic 0 to 100 macro gate, Claude news analysis per holding, alerts to Telegram, and a terminal-style Streamlit dashboard. It monitors; it never routes orders.

PythonClaude APISQLiteStreamlit
Working system

GenAI Document Assistant

Agentic retrieval over enterprise documents: planner, retriever, reasoner and validator agents produce grounded, cited answers with a grounded refusal when the source is missing. Runs fully local with no API key, or on Claude, Gemini or Ollama. Capstone for the Generative AI and ML program.

LlamaIndexChromaStreamlitDocker
Working system

TradingView MCP server

A Model Context Protocol server with 78 tools that lets an agent read and control a live TradingView desktop: chart state, indicator values, Pine Script compile and debug, replay, alerts, batch runs across symbols and timeframes.

MCPPine ScriptAutomation
Research

ORB 3-Step trading system

A deterministic intraday options system for IBKR paper trading with no LLM in the loop, a Pine Script prototype, and a backtest that acts as a GO or NO-GO gate before any execution code is written. Live ports are refused in code.

PythonIBKRBacktest gate
Research

Day Trading research lab

Eight strategy families tested to the same statistical standard, including 20,304 EMA-crossover variants. Seven rejected after Bonferroni and out-of-sample checks, one marginal. The negative results are the product.

MarketData.appStatisticsReports
Research

SPX Time Machine

An operator cockpit for the double calendar to risk-free iron condor transform on SPX, driven by live IV term structure from IBKR. Pure strategy math in one engine, a replica of a professional IV interface on top.

Node.jsIBKROptions math
Prototype

Great Circle

Finds the genuinely cheapest way to fly somewhere across cash fares, award points, split tickets and transfers, and shows the arithmetic. Live cash fares today; points connections next.

PythonLive faresTravel
Delivered

My AI Brain and Founder OS

A business operating system on Claude Code for non-technical founders: context files, connected apps, skills, workflows and memory, set up in a 25-minute onboarding without a terminal.

Claude CodeSkillsConnectors
Published · April 2026

Enterprise AI Infrastructure and Platform Advisory

The definitive edition of our infrastructure landscape: on-premises, cloud and hybrid platforms, LLM strategy, RAG architecture, GPU sizing and cost models for Fortune 500 and mid-market profiles, with an addendum on Nutanix and VAST Data.

AdvisoryInfrastructureTCO

Next step

Want one of these for your business?

Most of what is on this page started as a workflow someone was doing by hand. Tell us yours.

Book a 30-minute call → How we build agents