Gas Field Expansion Schedule Risk Analyzer with LangChain
L&T's contract award for Aramco's Jafurah fourth expansion phase involves critical timeline management across complex gas treatment facilities. Build a stateful schedule risk evaluation agent using LangChain and Llama 3.3 70B to detect critical path delays and resource bottlenecks. Maintain mean absolute error below 3 days on timeline prediction.
What you are building
The core problem, expected build, and operating context for this challenge.
Construct a LangGraph agent powered by Llama 3.3 70B to evaluate Primavera schedule exports and highlight critical path risks.
How work is evaluated
Evaluates prediction accuracy of task completion delays and critical path identification.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
critical_path_identification
Verifies correctly identified bottleneck task on critical path
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
delay_days_mae
Mean Absolute Error in predicted delay days compared to ground truth graph solver • target: 2.5 • range: 0-10
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Orchestrate stateful scheduling analysis flows using LangChain and LangGraph
Integrate Llama 3.3 70B model via local or inference APIs for reasoning tasks
Parse Primavera P6 / MS Project schedule network graphs and dependency chains
Predict float variance and schedule slip probabilities across engineering work packages
Reference links and supporting material
Network graphs for 30 EPC gas plant installation packages with Primavera P6 data fields.
Challenge setup
Evaluates prediction accuracy of task completion delays and critical path identification.
Challenge input
JSON graph of tasks with durational dependencies and current completion status
LangChain
Provides LangGraph for stateful multi-step graph processing.
Llama 3.3 70B
Delivers open-weights reasoning capability for complex industrial analysis.
Evaluated output
JSON containing predicted end date, total delay days, and critical tasks list
- Verifies correctly identified bottleneck task on critical path
- Mean Absolute Error in predicted delay days compared to ground truth graph solver • target: 2.5 • range: 0-10
- Benchmark: ScheduleRiskBench
- Delay Days Mae target: 2.5
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- Langchain · Required
- LangChain · Optional
- Llama 3.3 70B · Optional
- Langchain · Required
- LangChain · Optional
Evaluation contract
- critical_path_identification · Weight 1
- delay_days_mae · Weight 1
Metrics
- delay_days_mae · Target 2.5
Benchmarks
- ScheduleRiskBench
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON graph of tasks with durational dependencies and current completion status
Output: JSON containing predicted end date, total delay days, and critical tasks list
Execution harness: Python
Evaluation tools
- LangChain · Agent Workflow Framework
- Llama 3.3 70B · Language Model
Recipe freeze marker
No recipe freeze time is recorded.