Case studies
The work, on the record.
Recent engagements across NASA, Princeton, financial services, pharma, and energy. Cloud infrastructure, data platforms, AI/ML, compliance modernisation, MVP builds. The metrics and stories that come out of mission-driven work.
NASA / JPL · Mars mission cloud backend
Pixlise
The cloud backend for the Perseverance rover’s X-ray spectrometer.
PIXL analyses Martian rock chemistry live from the rover. We rebuilt the pipeline so a hypothesis and its answer can share a single sol, replacing an overnight batch with a streaming, fragment-tolerant system that recovers cleanly when a worker drops.
- 180×
- Faster pipelines
- 74%
- Platform TCO reduction
- 2023
- NASA Software of the Year, runner-up
Princeton University · Institutional research platform
The Accelerator
Research that took two years now takes three months.
A multi-source, medallion-architecture data platform for Princeton research on the information environment, with IRB-grade tiered access built in from the first commit rather than bolted on later. Live with 100+ beta users, continuing as the open-source Social Explorer.
- 4-8×
- Faster research cycles
- 100+
- Active beta users
- 12 mo
- Concept to live beta
Compliance · Global advisory (NDA)
Sanctions Compliance Platform
The best analysts in the business ran on a 240-tab spreadsheet.
We built the audit-ready platform that replaced it in three months, inception to deployment, against a hard licence-expiry deadline. It now runs across 20+ financial institutions, tenant setup takes days not weeks, and the AI layer is in production, added carefully.
- 3 mo
- Inception to deployment
- 20+
- Financial institutions
- 50
- Person delivery team
Fintech · Consilient
Consilient
Financial-crime detection deployed in a day, not weeks.
Consilient’s federated financial-crime models shipped on Apache Spark, and every bank deployment meant a bespoke cluster. Migrating the engine to Polars took deployment from weeks to roughly one day and made a single binary run inside any banking environment.
- ~1 day
- Deployment, was weeks
- Polars
- Replaced Apache Spark
- Any bank
- Environment compatibility
Pharma · MMIT
MMIT
Manual pharma data collection replaced by an autonomous extraction platform.
MMIT’s hybrid of automated scraping and human review couldn’t keep pace with the pharmaceutical intelligence market. We built the Databricks-native extraction platform that unified the pipeline, moved the manual work into models, and set the foundation for continuous ML improvement across the data estate.
- Databricks
- Cloud data platform
- Automated
- Ingestion pipeline
- Continuous
- ML improvement
Energy · Blackbrook Hydro
Blackbrook Hydro
Hydroelectric operations modernised, from disconnected sensors to AI-assisted control.
A hydro plant running on aging IT, disconnected sensors, and physical presence for routine tasks. We modernised the stack end to end, connected the sensor network, and used AI to accelerate PLC programming, bringing intelligent automation to a piece of critical infrastructure that had never had it.
- Unified
- Sensor visibility
- Automated
- Plant operations
- AI-assisted
- PLC programming
Ready to build your success story?
Partner with the team that delivered for NASA, Princeton, Consilient, and a global financial-crime advisory. A few lines about what you’re building is enough for Tom to write back honestly.
Talk to our team