# FieldLens

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Ievgen Voloshenko](https://cerebralvalley.ai/u/Sl1mCo)
- **GitHub:** https://github.com/Slimco86/FieldLens
- **Demo video:** https://youtu.be/h8wsHxbxMug
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4075

FieldLens: agentic reconstruction of oilfield assets from decades of historical records, powered by Meta Muse Spark.

Problem. Global energy demand keeps rising, now driven by electrification and AI data centres, and oil and gas still supply more than half of it. Producing that energy efficiently depends on understanding each asset, but an oilfield isn't one dataset; it's decades of them. Survey reports, completion logs, drilling data, well logs and daily production come from many contractors in many formats. The same well is named differently across sources, depths are measured along curving holes, completions change over time, and records conflict or leave gaps. Engineers can spend up to half their time finding and reconciling data, and wells end up managed on partial information.

Solution. In FieldLens, an engineer drops a field's raw files into a project, with no labels, schema or prior knowledge needed. A Meta Muse Spark agent turns them into an evidence-backed, explorable model of the field. Native parsers handle the structured formats. The agent reads what parsers can't, resolves which records describe the same wellbore, and designs each well's 3D asset. Deterministic code validates every proposal and computes all geometry, so every value traces back to its source page, table cell or line.

Features.
1. Upload with no preparation.
2. Automatic detection of data problems (duplicate names, missing locations, unit and depth-reference gaps, conflicts).
3. Agentic reconstruction: a five-stage run per wellbore with the cost shown up front and honest partial results.
4. Identity resolution with a human in the loop, backed by survey fingerprints and guard checks.
5. 3D reconstruction and visualization: wells at their surveyed locations, real trajectories, and casing, completion and perforated intervals at measured depth.
6. Traceability: one click opens the source PDF on the cited page.
7. Production, pressure and on-stream charts.
8. An agentic assistant that answers with exact dates, citations and inline charts.

## More from Meta Global AI Developer Hackathon

- [Reacknadam](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4071)
- [Nexel](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4072)
- [BEAXIS](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4073)
- [reefcensus](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4076)
- [GrocShoppers](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4077)
- [PocketDokani](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4078)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/4075. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt