Software developer · Fintech research tools · AI systems
I'm a Carnegie Mellon University student building full-stack products and AI-assisted workflows. I care about clear interfaces, deterministic logic and evidence that a reviewer can inspect.
Portfolio & case studies ↗ · Fintech portfolio · Public repositories · MOTIONX
An independent, AI-assisted fintech research workspace. Explore how distant cash flows affect value, how a hypothetical equity round changes ownership, and which assumptions explain a scenario difference. Keep operating cases with their reasoning and dated evidence.
Try the workspace ↗ · Engineering case study · Source
- See how much future is in the value. Value Horizon keeps the explicit forecast fixed and shows how the valuation changes when later terminal cash flows and residual value are removed.
- Explore an ownership constraint. Ownership Frontier, inside Financing Lens, connects a hypothetical primary round to dilution and the minimum stake retention you enter. Pre-money pricing, post-money pricing and operating enterprise value remain distinct.
- Explain and preserve an operating case. Move exact comparisons through Thesis Lab and the Model, then save the operating case and rationale to the Research Casebook. Casebook notes stay in the browser unless exported; financing links and exports retain their separate scope.
| Engineering question | Where to look |
|---|---|
| How is the calculation checked? | A shared browser/Node engine, real HTTP tests and separate Python parity checks. |
| Can research survive edits and failed saves? | The journey module preserves exact comparisons; storage tests exercise corrupt data, blocked storage and quota failures. |
| How are financing and dilution kept separate from operations? | The Financing Lens method explains the hypothetical equity bridge, round pricing, ownership retention and numeric limits. |
| What is reported evidence versus an assumption? | The contextual evidence method keeps source dates, definitions and modeling limits visible. |
Application 3.6 adds Ownership Frontier while retaining financial engine 3.0 and the 18 October 2026 research snapshot. Browser/API agreement establishes consistency; the independent Python implementation checks arithmetic. These checks do not establish forecast accuracy or investment returns. See the verification record and its scope.
Compose a review. Understand every decision. The Studio turns the original schema-change demo into an executable, typed workflow with evidence contracts, real step states, bounded retries and recovery. The hosted preview runs in the browser with synthetic tools; the separate local Node service adds durable SQLite checkpoints, revision checks and writer leases.
Try the source-outage scenario, restore its synthetic source and resume the same run. Inspect which evidence was reused, compare an additive alternative, and export the exact review. No paid inference, catalog mutation or SQL execution is enabled; deterministic components are not represented as an autonomous LLM team.
Verify what leaves the workflow. Review Inspector replays an exported review locally: evidence contracts, result-to-step agreement, deterministic conclusions and draft regeneration. It separates the recorded decision from current evidence freshness and guards JSON, Markdown and SQL exports. Files are not uploaded or restored as runnable state. Consistent is not authenticated; no execution is authorized.
Inspector method and limits · Semantic verifier · Negative-case tests · Three-minute portfolio walkthrough
Open Studio ↗ · Execution architecture · Recovery test · Verification and limits
| Project | What it demonstrates | Explore |
|---|---|---|
| LineageGuard Studio | A typed workflow composer and bounded execution runtime: source-bound evidence, cancellation/recovery, a separate SQLite local service and reproducible policy evaluations. | Studio · Source · Case study |
| MOTIONX Studio | A motion-first Next.js/TypeScript website, with a separate Python workflow for validated outreach drafts. | Source · Live site |
| Agentic Trading Research | A curated public architecture snapshot: model proposals, deterministic Python gates and evidence provenance. Research-only, with no automated order execution. | Public snapshot · Case study |
| ProofCast | Local narration and accessible cover generation, SHA-256 asset provenance and opt-in Backblaze B2 integration. | Source |
Also built: Blast Radius Daily, a deterministic TypeScript/Devvit puzzle; Teen Budgeting, a Node.js classroom demo with server-computed totals; and JUST-QR, a focused browser QR generator.
Languages and interfaces: Python, TypeScript, JavaScript, SQL, HTML/CSS, React, Next.js and Tailwind CSS.
Systems and verification: Node.js, REST APIs, JSON Schema, automated tests, GitHub Actions, Playwright, source provenance and human review boundaries.
I aim to make a project easy to try, its claims easy to inspect, and its failure states understandable. The portfolio checks and LineageGuard engineering review provide further entry points into the implementation.
The portfolio itself has an inspectable publication boundary: an exact public-file inventory, negative-case tests for accidental disclosure, and a byte-checked static release. The engineering evidence routes connect the visible demos to methods, code and tests—not unsupported usage or performance claims.
I'm interested in software engineering, AI engineering and product-focused internship opportunities. Visit the portfolio contact section or explore the public projects.
Projects include deployed interfaces, research snapshots, coursework and prototypes. Their READMEs explain what can be run and what remains out of scope. Company names in the valuation demo identify independent research subjects and do not imply affiliation or endorsement.