01 · Team Lead
Lansah
Owns ingestion, schemas, and reproducibility. Makes sure every score can be traced back to a row of source data.
● SAPTA · v0.1 · field-ready heuristic
Decide which pit to empty before the rain decides for you.
SAPTA scores rural pit latrines on groundwater-contamination risk and dispatches vacuum trucks under real capacity limits. Built for WASH teams in climate-stressed regions.
30 pits scored·10 dispatched·48h window·0.87 max priority
Upper East Region · synthetic dataset v0.1
The problem
“A single contaminated well can sicken a village faster than any truck can reach it.”
01 · LEACH
Pit latrines sit meters from hand-pump wells. Pathogens migrate through sandy vadose soils long before anyone notices a foul taste.
02 · RAINFALL
A single 40 mm event can raise the water table and flush latrine contents toward the nearest aquifer overnight.
03 · CAPACITY
WASH teams have a handful of vacuum trucks for an entire district. Routing is done from memory, phone calls, and gut feel.
Vacuum trucks can service ≈ 10 pits / 48h. A district has hundreds.
The solution
A priority index, not a guess.
One number per pit. Deterministic, explainable, and tuned by people who have walked the routes.
01 · INPUT
Telemetry + Static Data
Fill level, rainfall forecasts, soil class, distance to nearest well, pit condition.
02 · SCORE
SaniFlowCore (0–1)
Deterministic, auditable scoring function. Six weighted factors collapsed to a single risk index per pit.
03 · ROUTE
Greedy Dispatch
Capacity-aware selection of top-N pits within the 48h service window, with truck distance constraints.
01 · INPUT
Telemetry + Static Data
Fill level, rainfall forecasts, soil class, distance to nearest well, pit condition.
02 · SCORE
SaniFlowCore (0–1)
Deterministic, auditable scoring function. Six weighted factors collapsed to a single risk index per pit.
03 · ROUTE
Greedy Dispatch
Capacity-aware selection of top-N pits within the 48h service window, with truck distance constraints.
How it works
Six factors, one weighted score.
Each pit is reduced to a single value between 0 and 1. The dispatch engine reads only this number — but every number is traceable to its inputs.
- Fill levelf / f_max
- Closer to the brim, sooner overflow.
- Rainfall (48h)1 − e^(−r/15)
- Imminent precipitation amplifies leach risk.
- Distance to wellexp(−d/30)
- Proximity to drinking water raises urgency.
- Soil leach classk_soil ∈ [0,1]
- Sandy soils transmit pathogens faster than clay.
- Vadose depth1 / (1 + z)
- Thin unsaturated zone = shorter travel time.
- Pit conditionc ∈ {0.5, 1}
- Cracked liners short-circuit the calculus.
Note · v0.1
v0.1 is a calibrated heuristic — auditable, deterministic, hackathon-honest. v0.2 learns weights from labeled overflow events.
Live demo
From scores to routes.
The pipeline produces two artifacts a district WASH officer can actually open: priorities.csv ranked by risk, and an interactive map.html with the dispatch route already drawn.
No cloud login. No PowerBI license. Open the file, print the route, drive the truck.
Run it yourself
map.html · 30 pits · route_v0.1
$ python -m sapta.run --config config/upper_east.yaml [load] pits=30 wells=12 rainfall=48h [score] SaniFlowCore min=0.11 max=0.87 [dispatch] capacity=10 window=48h selected=10 [write] priorities.csv map.html ✓ SAPTA pipeline complete in 2.4s.
Hardware by Padigo & team · in development
The IoT layer
Sensing the pit, surviving the network.
The node assumes the worst: weak signal, dropped power, no NTP. Each field is captured locally and buffered until the next successful uplink.
- LEVEL
- ultrasonic distance to scum line
- POWER
- battery voltage + current draw
- LINK
- RSSI, retry count, last-seen
- TIME
- RTC fallback when network is dark
- UPLINK
- WiFi / GSM toggle, store-and-forward
Impact
Public health under climate stress.
Equitable sanitation operations.
Built to be handed to a district WASH officer on Monday.
SDG 6 · SDG 13 · WASH · Climate Resilience
Team
Four people, one pipeline.
02 · ML Dev & Frontend Dev
Abdulai Yorli Iddrisu
Builds SaniFlowCore scoring and the dispatch engine. Owns the path from raw telemetry to a deployable map.
03 · WASH Expert
Najat
Grounds the heuristic in real sanitation practice. Translates between district WASH officers and the model.
04 · IoT Engineer
Padigo
Designs the in-pit telemetry node — level, power, RTC, WiFi/GSM. Hardware that runs where the network doesn't.
Partner with us
Have pits to score, or data to share?
We're looking for district WASH partners, soil & hydrology datasets, and field-test sites for the IoT node.