Skip to content
Search jobs inside your AI agent. Claude, ChatGPT, Claude Code, Codex, Gemini, Cursor and Grok
Connect

InnoNext

Evidence‑Grounded Agronomy Copilot & MRV Metrics

Posted Updated Apply by 14 views

Description

Position overview:

Deliver the user-facing and integration layer: a grounded agronomy copilot that produces evidence packs (citations + retrieved snippets + quantitative outputs) and an MRV-style proxy dashboard from open data. This track also owns integration: wiring Tracks A–E into one runnable prototype with an integration test harness.


Problem statement (why this exists):

If an AI system cannot be audited, it will not be trusted in agriculture. This track builds a “show-your-work” assistant that connects recommendations to evidence and measurable proxies—while enforcing evaluation and guardrails against hallucination.


Data you will work with:

- Outputs from Tracks A–E (feature store, embeddings, models, causal effects, KG).

- Curated public technical sources for retrieval (papers, guidelines, standards) with strict citation requirements.

- Open data streams for MRV-style proxies; optional internal field data for validation (not required).


Key responsibilities (linked to the innovation project):

- Build a retrieval-augmented generation (RAG) workflow grounded in the KG (Track E) + curated technical sources.

- Implement guardrails and evaluation: groundedness, citation accuracy, hallucination testing, user-task metrics.

- Design MRV proxy metrics computable from open data (yield stability proxy, soil carbon proxy trend, water/input reduction proxy) with uncertainty.

- Create an integration harness that runs the end-to-end demo (containerised) and verifies module interfaces.


Why your expertise is crucial:

This role requires systems thinking: interface discipline, evaluation, and integration mindset—without it, the engine remains a set of disconnected notebooks.


Key outcomes for the Visiting Talent:

- An auditable, evidence-grounded assistant prototype + evaluation suite.

- Hands-on experience with integration and quality gates for AI systems.

- A demonstrable prototype suitable for deep-tech product narratives.


Stream:

- Deep-Tech Talents (research-focused).

Skills

Apply

See also