Led by Salim Bakri · Principal & AI Solution ArchitectBakri Consulting is a Canadian software consultancy led by Salim Bakri, specializing in AI architecture, cloud-native systems and full-stack delivery, backed by a network of senior engineers.
Services
From the first architecture sketch to production hardening, hands-on and accountable.
Turn LLM prototypes into reliable, secure products, from first architecture to production hardening.
Systems that scale with demand, stay observable and keep your cloud bill honest.
Hands-on delivery from MVP to production across web, mobile and desktop.
Senior judgment on demand: architecture reviews, due diligence and engineering leadership.
Not sure which one fits?
Most engagements start as a conversation. Tell me what you’re trying to achieve and we’ll shape the right approach together.
Approach
Pick the model that fits today and change it as your needs evolve.
1–2 weeks
Turn an idea or a problem into a clear architecture, risk list and roadmap, before you commit to a build.
Defined deliverables
A clear scope, timeline and outcome for a well-understood piece of work, from an AI feature to a new service.
Ongoing
A senior engineer inside your team on a retainer basis, shipping alongside you and raising the bar for everyone.
On demand
Architecture reviews, technical due diligence, mentoring and fractional-CTO guidance when you need judgment, not headcount.
We align on goals, constraints and what “done” looks like.
Architecture, trade-offs and a plan you can challenge.
Iterative builds with working software and honest status.
Docs, tests and knowledge transfer so you own the result.
Selected work
Client details are confidential, so engagements are anonymized. Products and projects I built myself are shown as-is.
Own product · AI engineering
A mobile learning app for senior engineers who want broad, adaptive coverage of topics outside their day-to-day stack. Designed, built and shipped end to end (currently in beta).
The problem
Adaptive quizzes and topic generation depend on LLM output, which is slow to arrive and sometimes malformed, a poor fit for a mobile experience that needs to feel instant.
Approach
Outcome
Own project · AI engineering
A production-grade MCP server that gives any compatible LLM client reliable access to a private document archive in many formats.
The problem
Teams want LLMs to answer from their own PDFs, Word files, spreadsheets and JSON, without locking into one vendor or stuffing everything into a prompt.
Approach
Outcome
Client engagement · Identity & security (anonymized)
Software for public identity and security programs serving Canadian and US government agencies, delivered as a cross-platform desktop application.
The problem
Each new product variant meant re-engineering workflows and hardware integrations from scratch, slowing delivery and multiplying defects.
Approach
Outcome
Client engagement · Health technology (anonymized)
A customer-facing web application and SDK integrating with mobile, cloud, QA and DevOps teams, including use on low-end devices.
The problem
An exposed encryption key, accumulated vulnerable dependencies, a heavy web bundle and slow deployments were putting security, user experience and release speed at risk.
Approach
Outcome
From the lab
More experiments in applied AI engineering.
Facing a similar problem?
Tell me about it. A short conversation is often enough to know whether and how I can help.
About
I’m a senior software engineer and AI solution architect based in the Greater Toronto Area. For a decade I’ve built software for banking, identity and security, health technology and the public sector, as an engineer, a technical lead and a mentor.
These days I help teams design and ship AI-enabled, cloud-native products: LLM integrations that hold up in production, serverless systems that scale, and codebases a team can actually maintain. I care about clear architecture, honest trade-offs and leaving your team stronger than I found it.
Every engagement is led and delivered hands-on by me. When a project calls for more hands, I bring in trusted senior engineers from my network, agreed with you up front.
FAQ
Something missing? Send a note and I’ll answer it directly.
Most work falls into four shapes: a short discovery sprint to scope and de-risk an idea, a fixed-scope project with clear deliverables, an embedded senior engineer or retainer alongside your team, or ongoing advisory (architecture reviews, technical due diligence, mentoring). If you're not sure which fits, book a call and we'll work it out together.
I work with a small number of clients at a time so each gets real senior attention, which means lead times vary. The fastest way to find out is a 30-minute discovery call: you'll leave knowing whether the timing and fit work, even if we don't proceed.
I'm based in the Greater Toronto Area and work in Eastern Time, which overlaps comfortably with teams across Canada and the US. Engagements are remote-first; on-site visits within the GTA can be arranged when they add value.
I do. Every engagement is led and delivered hands-on by me. For larger or time-boxed projects I bring in trusted senior engineers from my network, scoped and agreed with you up front, so you always know who is on your project and who is accountable.
By designing for it from day one: minimizing and redacting what reaches a model, keeping the architecture provider-agnostic so you choose the vendor and region, applying least-privilege access to data sources, and building monitoring and fallbacks around model output. We agree the data-handling rules for your context before any code is written.
Yes to NDAs, happily and before we discuss anything sensitive. Intellectual property terms are set out in the statement of work; the usual arrangement is that deliverables built for you belong to you.
I choose technology to fit the problem, not my résumé or anyone’s habits. I start from your constraints (your team’s skills, existing systems, scale, security and compliance needs, budget and timeline, and what you’ll have to maintain in three years), compare the realistic options and write down the trade-offs so the decision is transparent and easy to revisit. Sometimes that confirms the stack you had in mind; sometimes I’ll recommend something different and explain why. Where I have the most hands-on depth is TypeScript and Node.js, React and React Native, PostgreSQL, AWS (Lambda, CDK) and LLM tooling such as the Anthropic and OpenAI APIs, LangChain and the Model Context Protocol, but that’s where I’m strongest, not where I stop. If a project needs expertise I don’t have, I’ll say so and bring in the right person rather than stretch.
Contact
Tell me about the problem, or just book a call. No pitch deck required. I reply within one business day.
Prefer to talk it through?
Book a free 30-minute discovery call and pick a time that suits you.
Choose a time