Muhammad Talha Iqbal

Ex-Google DeepMind Engineer | Distributed Systems Architect

Based in: Lahore, PakistanMost recently: Sales & Business Development Lead, U.n.I — University Connectivity PlatformExperience: 7 years

Hire me

Verified top-tier engineer. Full portfolio: http://kanvis-verify.example/talha

Experience

Sales & Business Development Lead

U.n.I — University Connectivity PlatformLahore, Pakistanwork2023

Summary

Led sales and business development for a university-focused connectivity platform that scaled to 8,000+ active users within six months of launch.

What I did

  • Conducted on-ground qualitative user research across multiple campuses to identify friction points in the onboarding and peer-discovery loops.
  • Established partnership pipelines with university societies and student bodies, driving community-led acquisition without paid marketing spend.

Results

  • Restructured the early-experience flow based on in-person user research, reversing a decline in campus-level engagement during a period of 8,000+ user growth.
  • Prevented community churn by pivoting from aggressive acquisition to engagement-focused onboarding, ensuring long-term platform viability across multiple universities.
FlutterDjangoAWS

Founding Partner & Operations Lead

Cirkles— Event Discovery PlatformLahore, Pakistanwork2024 – Jun 2026
partnership relationships secured cold.
100+partnership relationships secured cold.

Summary

Joined as a founding partner with operational ownership across launch planning, partner outreach, and team coordination during pre-launch and early growth phases.

What I did

  • Built and maintained feedback loops with event organisers and early users; translated qualitative input into operational priorities for the product and engineering teams.
  • Coordinated cross-team workflows between design, engineering, and research collaborators to maintain release tempo through high-pressure build cycles.
  • Owned launch operations, partnership pipelines, and user-acquisition experiments; held the operating function.
  • Drove organisational decisions through structured analysis of team alignment, market signal, and product trajectory including the decision to exit in mid-2026.

Results

  • Personally built and secured a pipeline of 100+ client and event-organiser relationships during the pre-launch phase.
  • Analyzed qualitative market signals from 100+ partner conversations to identify friction in user behavioral change, leading to a structured strategic exit in mid-2026.

Data Analyst Intern

Systems LimitedPartnerLinq (US-facing Supply-Chain Product), Lahore, PakistanworkJul 2024 – Sep 2024

Summary

Worked within the PartnerLinq product team — a US-facing supply-chain integration platform — on workforce analytics and resource-allocation systems for the Operations function.

What I did

  • Built Power BI dashboards (DAX, SQL) tracking onboarding pipelines, team utilisation, and capacity allocation across delivery pods.
  • Contributed to an HR Resource Management System designed to improve allocation visibility and decision-making for product managers and team leads.
  • Exposed to workflow-orchestration concepts including DAG-based pipeline structuring and automation patterns used in operational data flows.
  • Translated raw operational data into structured reporting consumed directly by project coordinators across multiple agile pods.

Results

  • Developed predictive reporting models in Power BI that improved resource deployment accuracy across delivery pods.
  • Automated operational data flows, significantly reducing manual reporting time for project coordinators.
  • Optimized resource deployment and decision-making for 6 agile delivery pods, reducing manual reporting overhead.
Power BIDAXSQLDAG-based pipelines

Skills

Technical

PythonPython
3D modelling and visualisation
VS Code
Git/GitHubGit/GitHub
Local LLM deployment (Ollama)
Prompt-injection defence
Agentic-system architecture
Team leadership
HTMLHTML
CSSCSS
Power BI (DAX)
JavaScriptJavaScript
SQLSQL
FirebaseFirebase
Lang Chain
FigmaFigma
Wireframing
REST APIs
Pydantic
Agile/Scrum
User research
FlutterFlutter
AI-assisted dashboards
Data visualisation
Operational analytics
UI/UX design
Java
Penetration-testing fundamentals
MongoDBMongoDB
Workflow orchestration (DAG-based pipelines)
Node.jsNode.js
React.jsReact.js
DartDart
SQL ServerSQL Server
C++C++
Ethical hacking
AWS

Areas of expertise

Project management
Event management
Stakeholder coordination

Projects

Mitigation of Privilege Escalation in Autonomous AI Agents

Lead Researcher/Developerpersonal
Designing a multi-layer verification framework — the Federated Agentic Protocol — targeting prompt-injection-driven privilege escalation in LLM-based tool-calling environments. Implemented an isolated intent-verification layer using Phi-3 Mini (3.8B parameters) hosted locally via Ollama; verifier consumes only structured metadata (root intent, tool arguments, action-chain summary) and returns alignment scores via constrained JSON output with temperature pinned at 0.1 for reproducibility. Built a YAML-based declarative policy engine validated through Pydantic; enforces tool whitelists, parameter-level validators, scope constraints mapped to root-intent categories, and session-scoped rate limits at sub-5ms decision latency. Implemented a cryptographic hash-chain audit log providing tamper-evident accountability across all agent actions, enabling post-hoc verification of agent decision integrity. Evaluation against established agentic-security benchmarks in progress; framework packaged for open-source release as a Python SDK with a one-line developer integration via the IntentGate decorator pattern.
Phi-3 MiniOllamaJSONYAMLPydanticPython SDKIntentGate decorator patternCryptographic hash-chain
View the code

Education

Bachelor of Science

COMSATS University LahoreSoftware Engineering2023

Thesis. Mitigation of Privilege Escalation in Autonomous AI Agents

Coursework

Data Structures & AlgorithmsDatabase SystemsOperating SystemsComputer NetworksInformation SecurityPattern RecognitionHuman-Computer InteractionParallel & Distributed ComputingSoftware Project Management

Cambridge A-Levels and O-Levels

International School Lahore2023

Recognition

Awards

3rd Place, COLABS Pitch to Win

COLABS2025

Merit Scholarship

International School Lahore2023

Recipient of merit scholarship during A-Levels/O-Levels.

Silver Medal, Microsoft Hackathon

RAS Society

Winner, LUMS Psifi Science & Technology Festival

LUMS

Certifications

Machine Learning Specialization

Stanford University & DeepLearning.AI (Coursera)

Ethical Hacker

Cisco Networking Academy

Cybersecurity Learning Path

TryHackMe

UX Design Professional Certificate

Google (Coursera)

What drives the work

Motivations

  • Solving deep technical security challenges
  • Building scalable community infrastructure
  • Autonomous agent safety

Principles

  • Research-driven
  • Evidence-based decision making
  • Technical Integrity
Hire MuhammadReach out about a role, a contract or a conversation.For recruiters

Muhammad's twin is AI, it can make mistakes.

More people on Kanvis

Browse more peopleCreate your own Kanvis