CREDIT WORKBENCH
Leveraged Credit · Downside Engine
How this was built

What it is. An interactive credit workbench built around a simple question: how does a leveraged borrower behave when the world changes? It brings together the metrics lenders rely on, lets you stress the business through higher interest rates and weaker earnings, shows where covenant headroom disappears, and models how value flows through the capital structure in a downside scenario.

My role. I work in and around leveraged finance and private credit, and this project grew out of my own curiosity about how these pieces fit together. The credit framework, analytical workflow, assumptions and calculations are based on my understanding of the space. The implementation itself was built by directing AI coding tools, iterating on the logic, and validating the outputs against my own understanding and public market information. I'm not a software engineer by profession and don't claim to be — my contribution was defining what the engine should do, how it should think about a credit, and checking that the implementation reflected that intent. If you asked me why a calculation exists or what assumptions sit behind it, I should be able to explain it.

Where AI fits. One thing I wanted to explore was where AI genuinely adds value and where deterministic logic is still the better tool. Every financial output here is calculated directly from explicit assumptions — the numbers are deterministic, reproducible and traceable back to their inputs. The only place AI is used is the one-page investment summary: it receives the computed outputs and turns them into a short narrative, and never performs the calculations itself. That separation is intentional — I wanted the quantitative analysis to stay transparent while using AI only where natural language is genuinely helpful.

Domain + architecture: meBuild: AI tools, verifiedMath: deterministicSummary: AI narrates figures
Why I built this

I spend most of my time around leveraged finance and private credit, but over time I realised that understanding individual metrics wasn't the same as understanding how they interact. I could explain leverage, coverage or covenant headroom on their own, but I wanted to see how they moved together. What actually happens if interest rates rise? How quickly does a covenant become binding? How much of the reported leverage is driven by EBITDA adjustments? Which lenders are still protected if the business gets into trouble?

Those questions eventually turned into this project. I built it over a few weekends as a way to learn by building rather than just reading — changing assumptions and seeing the knock-on effects taught me far more than another spreadsheet or textbook could.

This isn't intended to be production software, and it certainly isn't a replacement for a full underwriting model. It's simply my attempt to understand leveraged credit in a more practical way while continuing to deepen my knowledge of the space. I'm also planning to pursue the FRM, and projects like this are one of the ways I try to connect theory with real-world decision making.

I'm sure there are assumptions here that could be challenged or improved. If you work in leveraged finance, private credit or restructuring and think I've misunderstood something, I'd genuinely appreciate the feedback. Conversations like that are one of the main reasons I wanted to share it.

Assumptions & simplifications

No model captures reality perfectly, and this one deliberately simplifies a number of areas. The goal was to build something transparent and educational rather than to replicate every detail of a credit agreement. Some of the key simplifications:

I believe models become more useful when their limitations are clear, so I've tried to make those trade-offs explicit rather than hide them.

Sources — Cloud Software Group (Citrix)

The featured example uses Cloud Software Group following the 2022 leveraged buyout and subsequent 2023 refinancing. It was chosen because it contains several characteristics that make it an interesting credit case — a layered capital structure, significant EBITDA adjustments, floating-rate exposure, covenant dynamics and refinancing risk.

All figures shown are illustrative, consolidated from public information and simplified for educational purposes. They should not be interpreted as current balances.

Primary references:

Some operating inputs — cash balances, capital expenditure, cash taxes, working-capital movements and scheduled amortisation — are reasonable illustrative estimates rather than directly sourced figures, and are labelled as estimates within the application.

The objective of the example is not to recreate the company's exact balance sheet at a point in time, but to demonstrate how the analytical framework behaves using a realistic capital structure that reflects situations commonly encountered in leveraged finance.

IN

The Credit

Interest & scheduled amortisation are computed from the tranches below, so they respond to the rate stress, hedging and revolver draw.
Reference-rate levels (editable, static — not live):
FX to USD (editable, static):
CREDIT WORKBENCHTear Sheet
—
Edits update live. Click Apply to confirm and refresh the readout.
—
—
Key risks
    Supports
      Full metric set—

      Leverage — size of debt i

      Gross debt / EBITDA i—
      Net debt / EBITDA i—
      Senior (1L) net / EBITDA i—
      Net nets out cash; senior counts only first-lien debt — the basis a springing first-lien covenant typically tests.

      Coverage — serviceability i

      Interest coverage i—
      Cash int. coverage i—
      FCCR i—

      Cash & liquidity i

      Free cash flow i—
      FCF conversion i—
      Total liquidity i—
      Liquidity runway i—
      Liquidity = cash + undrawn revolver. Runway = years of survival if the business is burning cash.

      Recovery & structure i

      LTV (debt / EV) i—
      Blended cash rate—
      Floating exposure—
      1L share of debt—
      LTV measures recovery risk; leverage measures earnings risk. Professionals watch both.
      CS

      Capital structure — how the debt stacks by seniority

      i
      Width is face value; the stack reads top-of-structure (paid first) to bottom (paid last). This is the order the recovery waterfall pours down.
      01

      Covenant cushion — how far EBITDA can fall before lenders gain control

      i
      The amber zone is the EBITDA decline absorbable before breach, shown in % and in dollars. The smaller cushion binds first.
      Both covenant cushions as EBITDA falls. Where a line crosses zero, that covenant breaches. The first to cross is the binding constraint.
      02

      Stress controls

      i
      Interest rates i+0bps
      ↳ of which floating debt hedged i%rate stress only bites the unhedged share
      EBITDA decline i−0%
      The "dash for cash": gross debt & cash both rise; interest rises as the commitment fee becomes a drawn rate.
      03

      Sensitivity grid — binding cushion across rate × earnings

      i
      Green safe · amber tightening · red breach. The view a credit committee reads first.
      04

      Recovery waterfall — who gets paid back in a default

      i
      At the current stress, distressed value = stressed EBITDA × × multiple. LTV .
      Senior secured (1L / unitranche) Second lien Unsecured / sub Impaired
      05

      Distress-masking lens — what the headline numbers hide

      i
      Adjusted vs. reported EBITDA
      Gross leverage (adjusted)—
      Gross leverage (reported)—
      Cash vs. total interest (PIK)
      Interest coverage (total)—
      Cash interest coverage—
      06

      Maturity wall — refinancing risk by year

      i
      07

      Observations — deterministic flags

      i
        08

        Current market context

        • Default activity in leveraged and private credit has risen off historic lows, with closer attention paid to PIK and deferred-interest structures as indicators of underlying strain.
        • Refinancing into a higher-rate, more selective market is a live pressure for borrowers with near-dated maturities, particularly floating-rate capital structures.
        • In software specifically, the pace of AI adoption has sharpened questions about the durability of recurring revenue and competitive moats — a key analytical input when underwriting SaaS credits.
        IS

        Investment summary

        Generated from the current assumptions. The engine above computes every figure deterministically; this turns those exact numbers into a short qualitative narrative. The numbers are calculated in code — the summary interprets them, it does not compute. The risk read is qualitative (High / Elevated / Moderate / Low), never a numeric probability.
        "I learn best by building. This project started as a way to answer questions I kept asking myself while working around leveraged finance, and gradually became a way to understand how lenders think about downside risk."Credit Workbench — credit logic, structure and validation by the builder; implementation built by directing AI coding tools. Deterministic engine; the investment summary is AI-generated from the computed figures. The featured capital structure (Cloud Software Group / Citrix) is illustrative, as of the 2022 LBO financing / 2023 refinancing, consolidated and simplified from public information (see Sources) — not current balances, not investment advice.