The gap between what a model costs on paper and what it costs on the device is where I work. Everything I claim, I measured.
I'm an engineer and researcher with an MS in Artificial Intelligence Engineering. My research concerns how vision models actually behave on embedded hardware, and why performance predicted on paper is an unreliable guide to what a device does under real conditions. Through Pesgit Engineering I translate this work into production edge systems; through Femtus Solutions I apply the same rigour to data and management reporting.
Edge AI · computer vision · embedded inference · measurement-led engineering
Performance Intelligence & Insight
Every system I build, whether for my ventures or directly for clients, is rooted in years of hands-on engineering and measured against real-world conditions, not demo environments. These are the principles that guide how I work.
Architectural Rigour
I design from the structure up; the underlying data model and architecture decide whether the result holds.
Execution That Ships
Strategy only matters if it ships. I anchor every engagement to a working output, tested against real data and real conditions, not just a review environment.
Reliability by Design
Systems have to perform under real conditions: real data, real edge cases, real operational pressure. I build for that from the start, and I'm honest about where the limits are.
Intelligence Frameworks
These are core capabilities I've developed and delivered, first as an engineer, and now through my ventures. Every framework listed below is built for real-world application.
Computer Vision & Edge AI
Through Pesgit Engineering: applied R&D in efficient real-time vision on constrained, offline-capable hardware, measured on the target device rather than estimated from specifications.
- On-Device Inference
- Measurement & Benchmarking
- Model Efficiency
Technical Advisory
Offered personally: advisory that bridges engineering reality and decision-making, covering architecture reviews, feasibility assessments, and technical roadmaps across data and AI systems.
- Architecture Reviews
- Feasibility Studies
- Technical Roadmaps
Data Analytics & Management Reporting
Through Femtus Solutions (London, UK): data acquisition and preparation, performance analysis, management and executive reporting, reporting automation, and project controls. I build the systems, not just advise.
- Data Acquisition & Preparation
- Performance Analysis
- Management & Executive Reporting
- Reporting Automation
- Project Controls
Edge AI Training
Through Pesgit Engineering: hands-on Edge AI training that takes engineering teams from a trained model to a benchmarked deployment on real silicon, with the measured evidence to prove it.
- Model to Silicon
- On-Device Benchmarking
- Measured Evidence
Selected Work
Real engagements and ongoing work across data systems, edge vision, and applied R&D.
Measuring What Compression Actually Costs
Research into how analytic cost models compare with measured device behaviour for vision model compression. The work measures what configurations actually cost on constrained hardware, under controlled conditions, rather than estimating from model structure.
KPI Framework for an Infrastructure Portfolio
Designed a KPI framework for a multi-million-euro infrastructure portfolio: aligned stakeholders, defined metric logic, and delivered executive dashboards and a phased roadmap.
Edge CV System: Assessment & Optimisation
Conducted an in-depth assessment of an edge computer-vision prototype, analysing inference design, latency characteristics, and optimisation pathways including hardware-accelerated inference approaches.
Vision-AI & Edge Perception
Through Pesgit Engineering: applied R&D in optical and thermal (EO/IR) imaging and efficient edge inference for industrial, agricultural, and infrastructure applications, taking hard perception problems from concept to field-credible prototype.
Why Clients Engage Me
These are the problems I see most often, and the reason clients reach out.
- Fragmented data and reporting with no single source of truth.
- Strategies that look good on paper but fail in practice.
- Edge/AI systems that work in a demo but break in the field.
- Large project schedules hiding underlying risk.
- Engineering and leadership talking past each other.
“Having built these systems from the ground up, from embedded circuits to executive dashboards, I understand where conventional approaches often fall short.”
Fawad Sarim
Claimed vs. Measured
Performance that only holds in a demonstration is not evidence. I work to the field standard, not the review standard.
The Demo Standard
- ✕Performance figures from structural estimates, not the target device
- ✕Validated in a lab, not under field conditions
- ✕Confidence without the measurement that earns it
- ✕Results that change when the environment does
The Field Standard
- ✓Latency and throughput measured on the target silicon
- ✓Systems validated under real operational conditions
- ✓Every performance claim tied to a number from the device
- ✓Methodology documented so results can be reproduced
Let's build something that actually works.
Data analytics, management reporting, project controls, and AI automation are delivered through Femtus Solutions (UK) and Pesgit Engineering (Pakistan). Computer vision and edge-AI R&D goes through Pesgit. For advisory, partnerships, and speaking — reach out directly.