I build AI-native systems that run real businesses.
Engineer, founder and technical advisor. I help companies in Africa, Canada and the US turn AI from a demo into dependable day-to-day operations.
Four ways to put a senior engineer on your problem.
Each engagement is scoped around an outcome, not hours. I take on a small number at a time so the work gets real attention.
Fractional CTO
Technical leadership without the full-time hire.
- Architecture, hiring bar and build-vs-buy decisions
- AI strategy that survives contact with budgets and reality
- Weekly leadership cadence with your team
AI engineering
Agents, voice, email and RAG that work in production.
- Tool-calling agents, MCP setups and RAG pipelines
- Voice AI, email automation and intent understanding
- Token optimization and the financial side of running AI
SaaS build partner
From idea to a system your team actually uses.
- Custom SaaS and business management systems through Ltiora
- Next.js, Node.js and AWS foundations built for reliability
- Designed around how the business really operates
Technical co-founder
Selective, long-term partnerships.
- For ventures where AI is central to the product
- Engineering, product and technical fundraising narrative
- Taken on rarely and only with strong alignment
Not sure which fits? Describe your problem or message me, and I'll tell you honestly whether I'm the right person.
What an AI-native business actually looks like.
Four layers. Most AI projects stall because one of them is missing. Hover or tap a layer to see what I build in it.
Assemble your own AI-native system.
Switch modules on and off and watch the stack take shape. I'll point out what's missing, the same way I would on a first call.
- Agents without a trusted data layer guess. Add data before autonomy.
- No evals means you can't tell when the agent gets worse. Add a judge.
Systems that run on their own, on real operations.
Momentum by Ltiora: the operating system for multi-store retail
Ltiora's flagship product puts point of sale, inventory and accounting in one system for multi-store retailers, especially those handling expiring stock such as pharmacies and grocers. The POS keeps selling offline and syncs when connectivity returns. Inventory spans locations with stock transfers and batch and expiry alerts, accounting runs in multiple currencies, and an AI layer reads supplier invoices and flags pricing, quantity and below-cost problems.
Ltiora: AI-native business systems
Ltiora builds management systems for businesses in the DRC and Uganda, now expanding into the US and Canada. The work spans custom SaaS applications, Voice AI, email automation and intent understanding, so day-to-day operations are run by systems rather than spreadsheets.
Data analysis and case management system
A full-stack system that gathers leads and builds case files for users. It answers client emails and calls automatically and manages meeting schedules. Built on Next.js, Node.js and AWS for reliability, security and a good user experience.
AI security and theft detection
An ML model paired with an LLM judge that identifies theft in live CCTV footage. It reaches an 80% confidence score on theft, concealment and suspicious activity, and is being put forward as a pilot for enterprise systems.
Digital twins and modelling research
A statechart-based approach that lets digital twins be built visually, and a domain-specific language that lets non-engineers build reinforcement learning models.
From IBM to Amazon to founding Ltiora.
Large-scale systems, fraud detection and customer service at big-tech scale, now applied to businesses where AI can change the day-to-day.
Senior Software Engineer · Senpilot
Jun 2026 to presentAI-native utilities management with tool calls and optimization. Analyzed data spanning millions of rows to refine the data model and lifted data ingestion alerting and overall data quality by 25%.
Founder & Technical CEO · Ltiora
Nov 2024 to presentAI-native business management systems for clients in the DRC and Uganda. Voice AI, email automation, intent understanding and custom SaaS.
Software Developer · Amazon
Oct 2022 to Jun 2024Built CRM software for Amazon Customer Service on GraphQL, Java and AWS, adding features that improved how agents help customers.
Software Developer · Verafin
May 2021 to Sep 2022Cloud infrastructure with Terraform and AWS to cut analysis latency. Worked on algorithms that track and analyze financial transactions and on fraud detection features.
Software Developer · Bluedrop Learning Networks
Mar 2020 to May 2021React front-ends, API and backend design for interactive, API-driven web applications.
Software Developer · IBM
Jan 2019 to Dec 2019Defect resolution and new features for IBM's analytics software with JavaScript and React. Led the Future Blue intern site programme.
Peer-reviewed, not just opinionated.
Two 2025 papers on making complex systems buildable by the people who understand the problem, not only the people who write the code.
Engineering Digital Twins with Statecharts: A Smart Home Application
Digital twins built visually from statecharts instead of hand-written glue code.
Read the paperTowards a Domain-Specific Modelling Environment for Reinforcement Learning
A framework that lets non-engineers assemble reinforcement learning models.
Read the paperHow I think about building with AI.
AI that earns its keep
Every model call has a cost. I design for the cheapest system that is reliably right, and I measure it.
Evals before features
If you cannot score an AI behaviour, you cannot ship it with confidence. I put evaluation in early.
Built for the way people really work
Systems fail at the human edge. I start from the operator on the ground, whether that is in Kinshasa, Kampala or Toronto.
Boring infrastructure, bold product
Reliable, secure foundations on AWS and Terraform so the interesting part, the product, can move fast.
A patent before a CS degree
In April 2017, before starting my computer science degree, I filed a patent for a semi-automatic dishwasher that cleans 500 to 600 dishes an hour. Hardware first, software after.
First place at the Ocean Innovation Hackathon (Techstars), 2018. IBM contributor to Call for Code, 2019.
Built where the operating reality is hardest.
Systems that work in low-bandwidth, high-stakes environments tend to be simple, resilient and honest. That discipline travels well.
DRC and Uganda
Ltiora's first clients: AI-native management systems for businesses, built around local operating reality.
Canada
Home base in Toronto, now taking Ltiora's approach to Canadian companies.
United States
Expanding Fractional CTO and AI engineering engagements for US startups and SMBs.
- ◆ Software developer and mentor with the Aga Khan Development Network since 2021, building support software for community programmes and mentoring businesses on using technology to improve their finances.
- ◆ Taught Python to more than 100 students from 5 countries through the Aga Khan Economic Planning Board in 2020.
Everything I've built, written and shipped.
The company, the product, the papers and the record, all one click away.
MODELS 2025 (EDT workshop)
Engineering Digital Twins with Statecharts: A Smart Home Application
Read the paperMODELSWARD 2025
Towards a Domain-Specific Modelling Environment for Reinforcement Learning
Read the paperWhat would it cost to run?
Most AI budgets are decided by a few numbers nobody checks. Move the sliders and see how volume, context and model choice change the bill, with real list prices.
List prices from Vercel AI Gateway list prices, as of 2026-10-07. A planning estimate: excludes embeddings, retries, tool calls and hosting. Real systems also cut cost with better prompts, smaller contexts and routing, which is most of what I do.
Skip the scroll. Put my AI to work.
Ask questions, find out honestly whether I'm the right fit, and get a draft plan you can send me. It knows my work and nothing else, and says so when it doesn't know.
Hi, I'm an AI assistant that answers questions about Sahil's work, services and background. I only use what's published on this site, and I'll say so when I don't know.
What would you like to know?