Options analytics and intelligence consultancy. We combine decades of derivatives expertise with AI to build tools and solutions for volatility traders. Purpose-built by practitioners, for practitioners.

You are an early senior engineering hire, and the first dedicated to shipping trader-facing features. Today the founder writes most of the product code -- tools, prompts, agent graph hops, sub-app UI, SQL -- alongside everything else. We need someone who can own the product-engineering surface end-to-end, make judgment calls without a tech lead in the room, and ship steadily against a full backlog. This is an IC role, not a manager role. **Your time will split roughly:** 80% building features end-to-end (backend + frontend + SQL), 10% writing regression tests and exercising LLM behavior, 10% triaging customer-reported bugs and translating them into fixes.

• **Add a small interactive UI affordance that drives an LLM-rewritten query.** A short word-search list, a click, a tool the LLM calls under the hood to splice the selection into running SQL, and the result re-rendering inline. Fullstack -- schema additions, a Python resolver, a tool definition, and a Dash UI -- in roughly a week. • **Add a lightweight pre-flight classifier to the agent.** Cheap rules cover the easy cases; a small LLM call covers the rest with prompt caching. Inline UI to act on the classifier's decision without blocking the user. Telemetry for every decision so we can tune. Latency budget in the few-hundred-millisecond range. A few days of active work. • **Add a new third-party data feed and expose it as a structured tool the agent can call.** Loader, two tables (header + chunks), tool schema, prompt-side guidance, regression cases, and the join keys that let downstream features pick it up. Around a week. • **Build a hybrid rules + LLM scoring pipeline.** Real-time stream of events, a small set of numeric signals composited into a score, items past a threshold enqueued for LLM enrichment under a daily budget, results landed into a new table and surfaced as a panel in one of the sub-apps. A handful of days across phasing (rules-only MVP first, LLM layer second, UI wire-up third). • **Harden the verification layer against a class of LLM failure mode.** Identify the failure pattern in real query logs, extend the verifier prompts (versioned in the database) so they can detect it, and wire the matching Python so the audit path actually acts on the signal. Small in code, high in leverage. Hours, not days. • **Fix a cross-stack edge case in pricing.** A real customer-reported bug whose fix touches the sub-app UI, the database layer, and possibly the upstream vendor feed; or load reference data into a cache to fill a gap the feed doesn't cover. Half a day to a day depending on path. • **Build a new tab in the main terminal on top of an existing engine.** Pick from a list, view a plan, execute it, render artifacts (tables, charts, narrative), plus a path that promotes a successful ad-hoc query into a reusable saved object. Backend, frontend, and LLM work all in one card. A couple of weeks. • **Move a heavy historical dataset off Postgres into columnar storage.** Partitioning scheme, within-file sort order, compression, manifests, a read helper that routes hot reads to Postgres and cold reads to DuckDB-on-blob, and a bit-exact reconciliation gate before the Postgres drop. A few weeks active.

• You have 5+ years of professional experience shipping product features end-to-end at a product company. You've owned features from spec to production, not just tickets, and you've been the senior IC on a small team before -- the person others ask for the call when there isn't an obvious right answer. • You are fluent in **Python**. Async patterns, type hints, packaging, testing. You can read someone else's 2000-line module and find the right place to make a change without breaking it. • You can build **frontend** -- HTML, CSS, JS, and at least one framework. We use **Dash / Plotly** heavily; React experience transfers cleanly and you'd ramp on Dash in a week. You care about UX latency, not just visual polish. • You write **PostgreSQL** comfortably -- joins, window functions, CTEs, JSONB. You can read `EXPLAIN ANALYZE`, you've authored indexes that mattered, and you know when to push compute into the DB vs the app. • You've worked with **LLM APIs** in production -- OpenAI, Anthropic, or equivalent. You know what prompt caching does, you've debugged a tool-call loop, you've shipped something where the LLM was a real component, not a demo. • You can ramp on a complex domain. Options pricing, vol surfaces, and the workflow language used by traders will be new. You're someone who reads the docs, asks sharp questions, and is productive on a derivatives codebase within a month. • You ship things. The bar here is taste and pace. **Nice to have:** Any agent framework in production. Anthropic SDK and prompt caching specifically. Dash / Plotly. Streaming UIs over WebSocket. Any prior exposure to options markets, derivatives, or financial data products. FastAPI, Quart, or similar async Python web frameworks. Experience writing structured tools (function-calling, OpenAPI) for LLMs to consume. You should NOT apply if you want to specialize on infra, deployment pipelines, or 24/7 operations -- that's a separate role we're hiring for, and the two complement each other. This seat ships the product.