AI for fixed income.
Applied Agents is a small advisory and build practice for fixed income. The work is two things, usually together: judgment about what AI is worth doing inside a regulated trading business, and the software to do it.
01 / The work
What the practice does
Advice - for this firm, at this size, what AI is worth adopting and what will quietly go wrong. The scarce input is no longer the models, which are commoditizing; it is the judgment to tell a real opportunity from vendor theater, and to size the response to the firm rather than to a brochure.
Build - working software, when a firm needs a result rather than another report: agents and assistants wired into a firm's own tools and data, workflow automation, and the data preparation that turns chats, dealer runs, and PDFs into something usable.
- Vendor-neutral Nothing to sell but judgment and build capacity. The recommendation is the one that fits the firm - including, sometimes, buying an existing tool rather than building anything.
- Capability transfer Every engagement is built to leave the firm's own team more capable, not more dependent on an outside advisor.
02 / Who it's for
Three sizes of buyer, one practice
The questions are the same kind of thing at any scale. What changes with size is the stakes - and how much a firm already knows it needs help.
The individual & operator
one desk
A trader, broker, analyst, or principal who wants to work faster without creating risk. Where AI fits the day, the tools to use, clear guardrails for what may and may not go into them - client data, confidential and material non-public information - and a sequenced plan for what to adopt first and how to tell it is working.
The growing firm
15-250 people
Past experimentation, wanting AI to become a governed capability without an enterprise budget or a dedicated AI team. An opportunity assessment ranked by value and risk, a readiness and data check, vendor-neutral build-versus-buy, pilots rolled into daily production, role-based training across front office, operations, and compliance, and governance sized to the firm.
The enterprise
20,000-100,000 people
Scaling AI across thousands of users and accountable to a board and regulators. The operating model and center of excellence, governance bodies and a federated policy, portfolio and intake management, enterprise AI risk integrated with model-risk management and internal audit, literacy at scale, and independent assurance - a second opinion, interim leadership through a transition, or AI due diligence on an acquisition.
03 / The specialty
Why fixed income
Large parts of fixed income still run on phone, chat, and spreadsheets - and that manual surface is where applied AI has the most leverage. Getting it right here means getting compliance and confidentiality right too: the regulatory reality of FINRA, SEC, and MSRB; data that is fragmented and largely unstructured; and model risk on pricing and valuation.
The use cases here can be built, not just advised on - parsing dealer runs and axes, assisting RFQ workflows, accelerating credit research, extracting terms from indentures and prospectuses, supporting evaluated pricing, building best-execution narratives, and surveilling communications.
Lead build targets
- Sales coverage & client outreach Surfacing what to say to which client, grounded in the firm's own inventory and history.
- Market color & morning commentary The easiest build and the clearest demonstration of value.
- Middle- & back-office liaison Trade breaks and reconciliation - high pull, low blast radius.
How anything gets built
- Embedded in the systems the desk already uses, not a replacement for them, and never positioned near price formation or execution.
- A person stays in the loop by design - every action is auditable, and nothing moves unless someone on the desk decides it should.
- If a tool stops working, the desk falls back to its normal manual process.
- The firm owns the result. The stack is Anthropic models on AWS.
04 / Background
A finance-native operator, not a technologist reaching toward finance
Behind the practice is roughly seven years front-office on Citi's Global Securitized Markets desk, ending as a vice president - including a real-time pricing, risk, and P&L platform for an $85 billion agency-MBS book spanning thirteen desks, delivered under OCC regulatory oversight.
That was followed by data and AI product work at streaming scale, and interim executive roles in the nonprofit sector. The career was spent productionizing and harmonizing disparate desk tools into working systems - the same work the practice now does for others.
- Credentials
- Master of Finance · CFA Level I · four AWS certifications · IAPP AI Governance Professional
- Domain
- Agency MBS, TBAs, specified pools, CMOs, prepayment modeling, OAS, FINRA TRACE, and SR 11-7 model risk
05 / How it starts
Nothing starts with a pitch
It starts with a plain conversation about where a firm actually is. Many are at zero, having never opened one of these tools - and that is a fine place to begin. The right thing to do depends entirely on the answer, so the first step is to ask rather than to assume.
Made concrete, the front door is a short, fixed-scope diagnostic: a four-hour "do I even need this?" review that returns a single page - what is real for the firm, what to ignore, and the one thing worth doing first. It is an honest offer: the answer can be that you don't need anything yet.
If it's worth ten minutes, the next step is a short call or an email.
Text or call +1 347 721 0838 (Signal, WhatsApp). Also on LinkedIn. References available on request.