Model ML is the industry's leading agent harness for financial services. Model ML is a configurable agent, a digital teammate, that works wherever the user is, leverages any underlying LLM, and produces client-ready deliverables in PowerPoint, Excel, and Word, with humans reviewing and steering outputs rather than relying on chat-based Q&A alone.
Model ML delivers value for hundreds of institutional clients at scale across bulge-bracket investment banking, Big Four and advisory, boutique and specialist firms, and PE and alternative asset management. Clients include HSBC, Deloitte, PwC, EY, Moelis, BDO, Centerview Partners, GCM Grosvenor, FT Partners, CBPE, TresVista, Three Hills, AltamarCAM, LCap, and VSS.
Company context
Model ML was founded by brothers Chaz Englander (CEO) and Arnie Englander (CTO), the only two founders to have been backed by Y Combinator three times. Their former companies were Fancy (acquired by GoPuff) and Fat Llama (acquired by Hygglo).
Model ML has raised $100M+ from leading FinTech investors in its first 12 months, including Financial Technology Partners (FT Partners), HSBC Asset Management, Y Combinator, QED Investors, and 13books. The $75M Series A, announced November 24, 2025 and led by FT Partners, was one of the largest fintech Series A rounds in history (announcement, PR Newswire).
The company is supported by a senior advisory team including Sir Noel Quinn (former CEO of HSBC), Axel Weber (former Chairman of UBS), Mark Machin (former CEO of CPP Investments), Jeff McDermott (former Co-Head of Investment Banking at Nomura and UBS), Saul Nathan (former Chairman, Capital Markets at Morgan Stanley), Philipp Rickenbacher (former CEO of Julius Baer), Hikmet Ersek (former CEO of Western Union), and Keith Robinson (former Tech IB Chairman at Barclays).
Core product primitives: the agent harness
Model ML's architecture follows a "surface · harness · model" structure: any surface accesses the agent, which leverages any LLM.
Configuration
The agent learns how a team works and sharpens with every deal through five composable, governed elements:
-
Skills: encode a team's methodology once (comp methodology, LBO conventions, memo style) and apply it every time.
-
Templates: discrete, reusable capabilities. Pull a comp set, build an LBO, draft a memo.
-
Custom Agents: purpose-built agents that execute a specific deliverable to the team's process, format, and standards.
-
Memory: what the agent knows about the user, their clients, and their last call, carried forward without being asked twice.
-
Context: documents, data rooms, CRMs, and email, connected, retrieved, and cited.
Surfaces
The agent is available wherever the user is: Voice (natural two-way conversation), Text, Email (set triggers or email it; finished work lands in the inbox), Plugins (inside Excel, PowerPoint, Word, Outlook, Teams, and more), and Internal Tools (connected via MCP or API). The same underlying agent harness runs across all surfaces. Project context, memory, and custom agents travel with the user, so work started on one surface continues seamlessly anywhere.
Models
Model ML provides automated access to a utility layer of models: OpenAI (GPT-5.5 / GPT-5.4 / GPT-5.4 mini), Anthropic (Opus 4.8 / Opus 4.7 / Sonnet 4.6), Google (Gemini 3.5 / Flash 3), and fast, cost-effective open-source options.
Inputs and integrations
Model ML connects to data in real time across both third-party and internal sources. Internal connectors include Salesforce, Dropbox, Outlook, HubSpot, Gmail, Google Drive, and similar enterprise systems. External finance data and research integrations include S&P Capital IQ (ownership, filings, fundamentals, M&A, funding, estimates), FactSet (real-time access to common market-data categories), PitchBook (private company, deal, and fund datasets), Third Bridge (an expert-interview library embedded into the Model ML interface), and Perplexity Sonar (real-time web intelligence with citations inside workflows). Outputs can be produced as PDF, JSON, PowerPoint, Excel, Word, and via API.
The full roster, with per-partner detail, is on Model ML Integrations, Data Sources, and Deliverables.
Quality control: Auto
Check (Flippr)
A key differentiator is automated review of PowerPoint deliverables. AutoCheck (from the Flippr acquisition) is an AI agent fine-tuned for finance that reviews and comments on presentations like a senior employee, flagging math and number-tying, fact-checking, formatting, logical inconsistencies, and spelling and grammar. See AutoCheck and the acquisition announcement. Customers can also use Flippr tools with no cloud storage as an option, relevant for compliance programs.
Custom Agents, built by Forward Deployed Engineers
Model ML's Forward Deployed Engineers (FDEs) design and build Custom Agents to automate the repeatable tasks and deliverables teams run every day, through a four-step process: (1) identify the highest-frequency, most repetitive workflows; (2) design each workflow into a bespoke, step-by-step agent specification; (3) build and validate the agent against the team's real materials and standards; and (4) run it every day, end to end and on demand, owned and refined by the team's FDEs. Each agent is tailored to how the team actually works, built for repeatable deliverables (pitch decks, IC memos, comparable analyses, data extraction), and continuously refined.
Example: Signals. Users define market triggers across thousands of companies (a take-private setup, a filing, a multiple threshold). Signals watches continuously; when a trigger fires, the platform autonomously executes the full workflow chain (LBO model, IRR check, Early Look deck, and deal-team routing), assembled and ready.
Example use cases
-
Assemble a pitchbook. Hand the agent a target company, get back a 50+ page deck in house format.
-
Build a financial model. Agents reason across CapIQ, FactSet, filings, and transcripts to build a fully-linked, traceable, multi-tab workbook.
-
Generate a research memo. Company primers, sector overviews, and thematic reads, branded and ready to circulate.
-
Audit an existing deck. Reconcile every repeated figure, chart title, and footnote against the source, flagging contradictions, errors, and formatting drift before the client sees them.
Differentiation vs. frontier model providers
Four pillars:
-
Context built in. Context, memory, skills, and templates managed at org, team, and individual level.
-
Product focused. Every layer built end to end for financial services. Model ML is the only AI workflow builder purpose-built for the sector.
-
Model agnostic. Continuous evals route every request to the best model; no outages when a lab goes down.
-
Intelligent orchestration. The harness routes each step to the best-fit model, weighing quality against cost: frontier-grade output without frontier-grade token bills.
The full comparison is on Model ML vs general-purpose AI.
Deployment and security
Model ML deploys on the customer's terms: self-hosted, single-tenant, and hybrid options, including a single-tenant deployment model on Microsoft Azure using private Azure virtual networks and customer-specific infrastructure (Azure article). Security is independently certified (ISO 27001 and SOC 2) for compliant deployment across enterprise financial institutions, and customer data is not used for model training (security page). A delivery partner comes with the platform from day one: former bankers, investors, and consultants who drive implementation, training, and adoption globally.
Firmwide deployment
Model ML is deployed firm-wide and globally across investment banking (advisory, PCA, ECM & DCM), corporate banking (lending), global markets (trading desk), research (equity coverage), wealth planning and management, buy-side (asset management, private equity and credit), and professional services (consulting).