Principal AI & Solutions ArchitectFounder & CEO, Thynkr SystemsLondon

Salim Chowdhury.

AI architect. Technology founder. Advancing governed intelligence.

I design and build intelligent software platforms and the AI infrastructure behind them, connecting advanced engineering with the practical work of running a business. Through Thynkr Systems, I lead product development and our research into governed AI improvement: systems that learn from evidence while people keep authority over consequential change.

Founder of Thynkr Systems · Building and leading products across AI, commerce, marketing, legal technology and financial intelligence · Software delivery since 1999

Salim Chowdhury, Founder and CEO of Thynkr Systems
Plate 01 PortraitSC-01
Role
Principal AI & Solutions Architect
Company
Thynkr Systems · Founder & CEO
Research
THYNKR RSI · Research & development
Practice
Software delivery since 1999

§01Architecture at a glance

Products, a shared intelligence layer, independent governance.

The architectural direction I lead across the Thynkr portfolio. Products keep their own data and business rules; Brain performs authorised AI work; WorQ governs authority and evidence. The improvement loop beneath them is research.

Governed intelligence architectureThynkr products at the top keep their own data and business rules. Brain, an existing foundation, performs authorised AI work with model routing and operational controls. WorQ, a live product, provides policy, approval and evidence. A dashed research box at the bottom shows the THYNKR RSI loop: observe, propose, test, approve, measure. PRODUCTS · OWN THEIR DATA, USERS AND BUSINESS RULES WorQPalletPOSTableSyncAIPROCRMGrowthPilotLexSalim AUTHORISED AI WORK POLICY · APPROVAL · EVIDENCE BrainSHARED AI INTELLIGENCE AND EXECUTION▸ Configured model routing▸ Operational controlsEXISTING FOUNDATION WorQGOVERNANCE, AUTHORITY, ACCOUNTABILITY▸ Agent and tool permissions, human approval▸ Reviewable evidenceLIVE PRODUCT THYNKR RSI · GOVERNED RECURSIVE SELF-IMPROVEMENTOBSERVE → PROPOSE → TEST → APPROVE → MEASURERESEARCH & DEVELOPMENT Governed intelligence architecture: products, Brain, WorQ and the THYNKR RSI research loop PRODUCTS · OWN THEIR DATAWorQ · PalletPOS · TableSyncAIPROCRM · GrowthPilot · LexSalim BrainSHARED AI INTELLIGENCE + EXECUTION▸ Model routing · controlsEXISTING WorQGOVERNANCE + ACCOUNTABILITY▸ Permissions · approval · evidenceLIVE THYNKR RSI · RESEARCHOBSERVE → PROPOSE → TEST→ APPROVE → MEASURE
Fig 1.0 Architectural direction · schematicIntegration depth varies by product
  • ProductsEach product retains responsibility for its own data, users and business rules.
  • Brain · existing foundationShared AI layer: configured model routing and operational controls. Wider orchestration is in development.
  • WorQ · live productGovernance for agent and tool activity: permissions, human approval and reviewable evidence.
  • THYNKR RSI · researchA target loop for evidence-led improvement. Not deployed as an autonomous engine.

§02Selected systems

Products I lead through Thynkr Systems.

Architecture, product direction and delivery, working with Thynkr Systems and its teams. These are company and team efforts; technology partnerships are identified separately.

All ventures & products →

§03Featured research

Research & development

THYNKR RSI — Governed Recursive Self-Improvement

Exploring how AI systems can observe their performance, identify improvements and evaluate better approaches through controlled experimentation and accountable human governance.

I lead the initiative’s architectural direction at Thynkr Systems: Brain as the shared intelligence and execution layer, WorQ as the independent governance and approval layer, and an evidence loop in which a system that proposes a change can never approve its own authority.

Status

A research and development direction, not a product or service. Brain and WorQ foundations exist; an autonomous self-improvement engine is not deployed.

Target improvement loop

  1. ObserveBaseline permitted operational signals.
  2. ProposeExplain an opportunity, its risks and evidence needed.
  3. TestCompare candidates in isolated, reproducible experiments.
  4. ApproveMaterial change goes to accountable people.
  5. MeasureCompare outcomes with the baseline; revert what fails.

Measure → Observe · verified results inform the next cycle

§04Technical expertise

Where the work sits.

Each area is tied to work I lead or have delivered. Research interests are kept separate on the research pages.

01

AI systems architecture

Architecture for governance-first AI operations across the Thynkr portfolio, from product intent to production design.

Thynkr Systems · 2023—present

02

Multi-model infrastructure

Directing Brain, Thynkr’s shared AI layer: configured model routes and operational controls across products.

Brain · existing foundation

03

Agentic software systems

Products where people and AI agents work on shared records within a common permissions model.

WorQ · AIPROCRM

04

AI governance

Defining what AI can access, decide and do, with human approval and evidence that can be reviewed later.

WorQ · AI Governance Readiness Review

05

Cloud & distributed systems

Secure digital services, platform boundaries and transition paths that connect infrastructure choices with the products they support.

Independent consulting · Klixspot · Nexero

06

SaaS platform engineering

Multi-tenant platform engineering for software products, with tenant boundaries, permissions and release controls designed in from the start.

Thynkr portfolio

07

Operational platforms & payments

Counter sales, credit accounts, invoices and statements; restaurant POS, kitchen and staff flows. Electronic payment collection in TableSync is in development.

PalletPOS · TableSync

08

Modernisation & recovery

Making dependencies visible, defining transition states and recovering stalled delivery for data-intensive systems.

Platform Recovery Assessment · consulting since 2019

§05Experience

Strategy informed by hands-on delivery.

A career spanning software development, programme leadership, data, cloud and platform architecture. The CV provides the fuller chronology.

  1. 2023—present

    Founder & Chief Architect · Thynkr Systems

    Product direction, governance-first AI operations and multi-tenant platform engineering across the Thynkr portfolio.

    Founder and CEO; leads THYNKR RSI research direction.

  2. 2024—present

    Data, AI & Cloud Consultant · Independent

    AI adoption, cloud and data modernisation, integration and delivery recovery.

    Fixed-scope reviews and fractional architecture.

  3. 2021—present

    Independent product architecture

    PalletPOS from 2021; Nexero from 2022.

    PalletPOS is now a Thynkr Systems product.

  4. 2019—2023

    Data & Cloud Consultant · Klixspot Ltd

    Modernisation, integration and delivery across data-intensive business systems.

    Data and cloud consulting.

  5. 2011—2018

    Senior Project Manager / Consultant · Westech

    Project and programme leadership.

    Earlier consulting, project-management and development roles from 1999.

§06Engagements

Three ways to start.

Fixed-scope engagements with a clear outcome, so you know what you are buying before we begin.

AI Governance Readiness Review

For teams adopting AI agents, coding assistants or automated workflows. I assess where AI can act, what it can touch and what evidence you keep, and deliver a prioritised control plan.

How the review runs →

Platform Recovery Assessment

For stalled or over-budget programmes. I map the dependencies, identify the real blockers and agree a delivery sequence your team can execute.

How the assessment runs →

Fractional Chief Architect

Ongoing architecture leadership for founders and scale-ups: decisions, reviews and delivery oversight, without a full-time hire.

How it works →

§07Perspectives

Writing in preparation.

No articles are published yet. These are the topics I intend to write about; nothing below is a published paper or talk.

  • 01Governed recursive self-improvementPlanned
  • 02Multi-model orchestrationPlanned
  • 03AI agent assurancePlanned
  • 04Practical AI evaluationPlanned
  • 05Production engineering lessonsPlanned

Start a conversation

What needs to move forward?

A platform decision, a stalled programme, a new product or a research collaboration. Thirty minutes is usually enough to know whether I can help.