Mann Patel Forward Lead AI Strategist

Mann Patel

Three years delivering for global banking, capital markets and insurance institutions. Now I design and ship the agentic systems that turn raw data and rough ideas into finished, reliable output: live products on their own domains, markets tooling, automation, creative production, and the design systems that make it all look intentional.

Financial Services Delivery Orchestration and Automation AI Assisted Markets Analysis Live Product Launches Creative Production Pipelines Design and Document Systems
Live
Midnight Jot ↗midnightjot.com
A journaling PWA on its own domain: offline first capture, account sync across devices, and a corpus structured for analysis from the first keystroke.
Fablewick ↗Published
An open source storybook library: 11 illustrated children's books, each in 3 reading levels and 5 languages, with character consistency held across every page by a repeatable generation pipeline.
Ramkabir Fablewick ↗Heritage Edition
A specialized heritage collection on the Fablewick engine: five books planned, written, illustrated, and published in a single day, holding family and faith on the page.
Motion Backdrops ↗In Use at Live Events
Animated stage backdrops generated and looped in house, delivered for real ceremonies and projected behind the moment itself.
Background

Where This Comes From

Before the AI work there were three years inside regulated financial services, delivering for institutions where the controls matter as much as the build. The two halves are the point.

Read the full résumé →

AccentureMay 2023 to Feb 2026
Business Consultant

Owned the deal workflow for a book of banking, capital markets and insurance institutions: 200+ concurrent opportunities against $50M+ in annual volume, as the single point between the client and delivery. Drove platform adoption past launch for the Home Loans division of a top 5 bank in New York City, writing the user stories, running the sprint cadence, and reporting business unit value on rollout so the people funding it could see what had changed. Redesigned how work entered and moved through the delivery queue across concurrent engagements, cutting backlog 30 percent in six months and raising sprint velocity.

Corient (formerly RegentAtlantic)2022
Wealth Management

Supported 20+ high net worth clients with $5M+ portfolios across retirement, tax, investment and estate strategy, and ran due diligence on 90+ investment funds against prospectus alignment, benchmark performance and risk parameters.

Fair360 (formerly DiversityInc)Jan 2020 to Jun 2021
Data Analytics

Built a JavaScript application over 20 years of Top 50 data (2001 to 2020), cutting the analysis cycle by 80 percent. The first thing here that was code solving a data problem, three years before any of the AI work.

Rutgers Business School2023
Bachelor of Science in Finance

Dean's List 2019 to 2023, 3.81 GPA. Rutgers Future Scholars Dream Team since 2014 as Scholar, then Treasurer, Director of Community Service and Ambassador Alumni, mentoring 2,000 first generation and financially disadvantaged students.

What I Do

Capabilities

Two halves that work together: the consulting craft of delivering on regulated platforms, and the engineering to automate the work underneath. AI does the deterministic heavy lifting; I own the judgment, taste, and structure.

Delivery

Programs and Delivery

Business analysis and scrum delivery on regulated platforms: user stories, sprint cadence, change management, and business unit value reporting on rollout and adoption.

Advisory

Client and Wealth Advisory

CRM pipeline management at scale, RFI and RFP proposal work, and wealth management grounding across retirement, tax, investment and estate strategy plus fund due diligence.

Systems

Agentic Orchestration

Multi layer workflows, self improving scheduled loops, a persistent external memory system that keeps state coherent across sessions, and multiple coding agents working one repository under a written handoff protocol.

Context

Prompt and Context Engineering

Designing the instructions, guardrails, and memory that make AI output reliable, on brand, and repeatable.

Markets

AI Assisted Analysis

Reading macro environments, sectors, and futures instruments to gauge the quality of conditions, structured and layered, on live data pulled from the Yahoo Finance API and a Robinhood MCP server.

Build

Full Stack Delivery

Next.js and React front ends, offline first PWAs with account sync on Supabase, Python data and automation, HTML to PDF pipelines, DNS and custom domains, shipped, verified, and deployed.

Create

Generative Production

Image, video, and document generation wired into repeatable pipelines with consistent characters, style, and QC.

Connect

Connectors and Automation

Gmail, Drive and Calendar workflows plus market data MCP servers, built as reusable playbooks, one runs on the next account by changing a label.

Selected Work

What I've Built

Grouped by domain. Each is something I designed, built, and shipped. Most are live or in daily use.

1AI Assisted Financial Markets Tool
A decision support system for reading market conditions before taking risk.
Multi Layer Market ReadLive
Analyzes the macro environment, sector rotation, and the futures instruments that lead it, layering structure, macro context, and order flow into a single graded read of how favorable conditions are for putting risk on.
Self Rendered Charts and a Grading FrameworkPython · Next.js
Pulls OHLCV bars from the Yahoo Finance API and renders its own candles rather than screenshotting a terminal, applies a consistent checklist based rubric, and logs every read with its outcome, so the framework can be measured and improved, not just trusted.
Live Market Data LayerYahoo Finance · Robinhood MCP
Two sources wired behind one interface: Yahoo Finance for historical and intraday bars, and a Robinhood MCP server for real time quotes, option chains, fundamentals, technical indicators and screener runs. Every figure is reconciled against the source export before it reaches a chart or a document.
2Agentic Systems and Automation
The infrastructure that lets AI run real, repeatable work with the human in control.
The Exocortex, a Persistent AI Operating SystemDaily Driver
An external memory and meta agent system: domain agents, self improving scheduled loops, and a witness layer that surfaces and records but never decides for me. It keeps state coherent across sessions and compounds judgment instead of rebuilding it each time.
Connectors and MCP ServersGmail · Drive · Calendar · Robinhood
Email, Drive and Calendar automations built once as playbooks that run on a new account by changing a label, not rewiring the workflow. The same pattern drives market data MCP servers, so a quote, a fundamental and a screener result arrive through one contract.
Local Speech to Text ToolFaster Whisper
A self hosted dictation utility (push to talk, tiered cleanup), a from scratch local clone of a paid SaaS, built to run offline.
3Creative AI Production
Generative pipelines that ship finished, consistent creative, at speed.
Fablewick, Open Source Storybook Library ↗Published
11 illustrated children's books, each in 3 reading levels and 5 languages, with 66 hand styled watercolor illustrations, character consistency held across every page via a repeatable generation pipeline. A 5 book heritage collection was later planned, written, illustrated, and published in a single day on the same engine.
Tribute Films and Motion PiecesHiggsfield · ffmpeg · Remotion
Photo to film tributes and code driven motion graphic explainers: face centered montages, lyric films, and programmatic videos built from in house pipelines.
4Design and Document Systems
A house style and engines that make every deliverable look intentional.
A Codified Design SystemReused Everywhere
One palette, one type system, one set of rules, applied across PDFs, decks, and web so output is consistent and premium by default, not per project guesswork.
Report and Document EnginesReportLab · HTML to PDF
Programmatic PDF pipelines (ReportLab with matplotlib, and Playwright HTML to PDF) that reconcile to a source export first, then lay out: repeatable monthly reports and data documents.