FilingLens design candidates

Best for: presenting FilingLens as a serious financial research cockpit.

Structure

Console hero, KPI strip, evidence table, answer contract, technical stack.

First screen

Dark console with sample query, metric cards, API/mock status, and cited answer preview.

Middle

Financial facts, SEC text evidence, risk themes, valuation boundary modules.

Demo

Static NVIDIA query with mocked tool calls and final cited answer.

Best for: emphasizing auditability, evidence lineage, and runtime checks.

Structure

Trace hero, stage timeline, evidence matrix, contract outcome, limitations.

First screen

Trace dashboard with run state, required answer parts, and contract status.

Middle

Tabs for ResearchPlan, EvidencePacket, Dimensions, Citations, Contract.

Demo

Mock trace inspector showing how each sentence maps to [N] and [T] evidence.

Best for: showing the product as a practical research workspace.

Structure

Query panel, parsed intent, evidence needs, answer preview, trace drawer.

First screen

Split workspace where a natural-language question becomes structured analysis state.

Middle

Saved query examples for cash flow, revenue growth, risk, and comparison.

Demo

Static query editor with mocked intent chips and answer outline.

Best for: explaining the full agent workflow without overwhelming the reader.

Structure

Hero pipeline, stage-by-stage walkthrough, architecture, eval section, demo.

First screen

Large horizontal flow from query to validated cited answer.

Middle

Seven stage cards: classify, plan, execute, evaluate, generate, check, finalize.

Demo

One sample query advances through static checkpoint panels.

Best for: making source grounding the main visual identity.

Structure

Citation hero, evidence library, answer assembly, contract checks, limitations.

First screen

Citations appear before the answer, showing evidence as the product foundation.

Middle

Numeric evidence, text evidence, inference, and missing-evidence lanes.

Demo

Mock answer with highlighted [N] and [T] citations connected to source cards.

Best for: showing tool orchestration and system design to engineers.

Structure

System-map hero, tool protocol, data stores, state machine, demo trace.

First screen

Central LangGraph hub connected to tools, stores, and validation layers.

Middle

ToolResult schema mock, provenance, latency, errors, and deterministic calculations.

Demo

Tool-call transcript for a revenue growth question with mocked latencies.

Best for: foregrounding reliability, testing, and measurable evaluation.

Structure

Reliability hero, eval caveats, test coverage, probe board, architecture appendix.

First screen

Quality dashboard with checks, counts, and small-sample caveat.

Middle

6-case release signal, 300-case demo probe, 78 test files, failure categories.

Demo

Mock contract runner showing pass, repair, and blocked outcomes.

Best for: making the SEC research angle concrete and memorable.

Structure

Terminal hero, corpus stats, filing sections, retrieval examples, cited answer.

First screen

Command-like research terminal querying filings and structured facts.

Middle

10-K/10-Q corpus, section-aware chunks, financial facts, price history.

Demo

Static terminal transcript for a revenue-driver question.

Best for: explaining why this is not a generic RAG demo.

Structure

Problem hero, before/after answer, contract repair, evidence trace, metrics.

First screen

Side-by-side weak answer vs grounded answer with visual quality checks.

Middle

Missing evidence, unsupported conclusion, repaired answer, final contract status.

Demo

Mock comparison for advice-like AMZN query and safe rewrite.

Best for: making the stack and component boundaries the story.

Structure

Architecture hero, data flow, subsystem sections, demo inset, reliability.

First screen

Layered system diagram from UI to API to agent to data stores.

Middle

FastAPI endpoints, LangGraph nodes, DuckDB tables, Chroma retrieval.

Demo

Small query console embedded beside architecture callouts.

Best for: fast scanning by recruiters and interviewers.

Structure

Short hero, three contribution cards, demo preview, stack, outcomes.

First screen

Plain-language value statement plus three concrete engineering highlights.

Middle

What I built, why it matters, how it was evaluated.

Demo

One polished mock answer, minimal controls, clear citations.

Best for: readers who want implementation details and system contracts.

Structure

Technical hero, state schema, graph nodes, tool protocol, validation.

First screen

Documentation-like layout with core abstractions and contract output.

Middle

AgentState, ToolResult, EvidencePacket, DimensionStatusMap, repair logic.

Demo

Trace tabs plus compact pseudo-code snippets and mock JSON outputs.

Best for: showing the full reasoning loop in one first-screen composition.

Structure

Three-pane hero, expanded pane sections, demo query bank, tech notes.

First screen

Query, evidence, and answer are visible at the same time.

Middle

Each pane becomes its own section: parser, evidence matrix, synthesis.

Demo

Static three-pane mock app for comparison queries.

Best for: a compact, technical, high-contrast project identity.

Structure

Dark hero, live-run monitor, evidence cards, trace audit, project notes.

First screen

Run monitor with mocked node timings and contract outcome.

Middle

Console logs, tool payloads, citation checks, final answer block.

Demo

Mock terminal output for an end-to-end run.

Best for: a polished portfolio page that reads like a concise case study.

Structure

Editorial hero, problem, approach, demo figure, results, technical appendix.

First screen

Typography-led opening with a quiet system figure beside the summary.

Middle

Problem framing, design decisions, evidence contract, evaluation caveats.

Demo

Embedded static figure showing query, evidence, answer, and limitations.

Best for: emphasizing local-first architecture and model-provider flexibility.

Structure

Local stack hero, serving boundary, data stores, agent graph, demo.

First screen

Topology showing local UI, API, compatible LLM endpoint, and local stores.

Middle

Why deterministic tools own calculations while the LLM handles bounded language tasks.

Demo

Mock run that labels model, tools, and validation as separate responsibilities.

Best for: highlighting safety, grounding, and non-advisory financial boundaries.

Structure

Contract hero, advice-like query rewrite, citation checks, limitation states.

First screen

AnswerContract checklist next to a risky query transformed into a safe analysis task.

Middle

Numeric grounding, citation validation, DimensionStatusMap, forbidden claims.

Demo

Mock blocked or repaired answer with visible rule outcomes.

Best for: foregrounding the financial-analysis framework instead of the software stack.

Structure

Analysis framework hero, dimension sections, evidence examples, demo answer.

First screen

Seven financial-analysis dimensions with a sample query mapped to required evidence.

Middle

Revenue quality, cash-flow quality, profitability, balance sheet, moat risk, valuation boundary.

Demo

NVIDIA or AMZN workflow with metric table and risk citations.

Best for: a sober, research-oriented project presentation.

Structure

Abstract, contributions, method, system, results, limitations, demo figure.

First screen

Paper-like abstract and contribution table, with restrained typography.

Middle

Method and evaluation sections with clear caveats for sample size and demo probes.

Demo

Figure-style mock of query, evidence, answer, and validation.

Best for: making the preview feel alive while staying fully static.

Structure

Interactive mock hero, query tabs, response panel, evidence drawer, architecture below.

First screen

Clickable local query tabs that update mock intent and answer text only in the browser.

Middle

Selected query expands into evidence needs, tool calls, citations, and contract checks.

Demo

Static JavaScript switches between NVIDIA, Amazon, and Apple mock cases.