I build enterprise assistants, voice agents and multi-agent workflows around shared runtimes and harnesses. The runtime manages context, durable state, capability loading, tool and code execution, recovery and approval gates for sensitive actions. Voice systems add realtime transport, conversation state and interruption handling.
Enterprise AIIdentity & SecurityDistributed Systems
I design and build
production systems
for security, reliability and scale.
I'm Hammad Abbasi, a software architect with 16+ years combining technical advisory with hands-on engineering. I build AI runtimes and protocols, identity platforms, security controls, and high-performance distributed systems.
I've spoken at the European Identity & Cloud Conference in 2024 and 2025, and at Medium Day 2026. My AI security work has also been cited by RAND and the Journal of Advanced Military Studies.
Agent Runtimes & HarnessesOrchestration · Durable State · Voice & Realtime
Retrieval & Search SystemsAST/Metadata-Aware Retrieval · RRF
AI Security & IdentityRed Teaming · Secure Tool Execution
Distributed Systems & ProtocolsgRPC · HTTP/3/QUIC · Event-Driven Systems
Built enterprise platforms used by leading organizations
Production Experience
Systems built for real production workloads
I've designed and built systems across AI, identity, last-mile delivery marketplaces, multi-tenant SaaS and high-volume document platforms — including real-time geospatial matching, event-driven services, identity infrastructure, agent runtimes and APIs handling millions of requests each month.
16+ yearsArchitecture + hands-on engineering
98% reductionHallucinations in evaluated AI responses
80% lower fixed overheadIris controlled benchmark
1.6M usersIdentity and SSO platform
<100 msReal-time geospatial matching
Products
Products I've built and operated
Independent products across AI infrastructure, consumer AI and multi-tenant SaaS, each taken from concept into live use.
01Featured product
Iris
CreatorLocal control and policy enforcement for Claude Code
A local control plane between Claude Code and Anthropic. Exposes context and tool schemas, controls available capabilities, evaluates tool calls against deterministic policy and records execution.
28K5.6K tokens
Fixed request overhead after removing unused tool schemas.
- Context visibilityInspect model traffic, context and tool schemas.
- Tool controlControl available capabilities and tools.
- Deterministic policyEvaluate tool calls against defined policy.
- Execution recordRecord and audit what actually ran.
Research & open source
Research and open source
I build and publish systems around agent runtimes, continuous authorization, cross-system coordination and deterministic search.
Research paper
Outcome Coordination for AI Agents
Coordination across independently governed systems.
A protocol architecture for expressing objectives, offers, commitments and evidence across organizational boundaries while leaving execution under the control of each participating system.
01
View Lattice Lattice
Agent runtime and coordination framework.
A runtime for agent systems with durable state, tool execution, authorization and recovery outside the model loop.
Auto-cycling projects ·
Experience
Architecture backed by production delivery
16+ years building systems across AI, identity, last-mile delivery, SaaS and document platforms, with roles spanning architecture, hands-on engineering and technical leadership.
16+Years
1M+Daily operations
1.6MIdentity users
5M+Monthly API requests
EmpowerID
Sep 2023 – Present · Remote, United StatesAI Software Architect
Worked across EmpowerID’s AI and identity platforms, spanning enterprise agents, voice agents, retrieval, secure tool execution, delegated authorization and distributed services.
- Built EmpowerIQ Docs, using query routing and metadata-aware retrieval; reduced hallucinations by 98% in evaluated responses.
- Built a shared agent harness for context management, durable state, recovery, approvals and capability loading.
- Designed identity and authorization infrastructure covering workload identity, token exchange, delegated authority, policy enforcement and audit.
- Built secure capability and tool-execution controls around authenticated production actions.
WappGPT
May 2022 – PresentFounder
Built and operate a production AI assistant around WhatsApp with conversational memory, natural-language recall and durable reminder orchestration.
- Built acknowledgement-driven workflows that preserve state across channels.
- Unacknowledged reminders can escalate from WhatsApp to SMS to phone while retaining context.
- Responsible for the AI, orchestration, messaging, billing and production infrastructure.
Point Pickup Technologies
Feb 2022 – Apr 2023 · Remote, United StatesLead Architect
Led architecture for a last-mile delivery marketplace serving retailers including Walmart, Kroger, Dollar General and 7-Eleven.
- Scaled from 250K to more than 1M orders per day.
- Increased delivery success from 92% to 97% and reduced cancellations from 11% to 4%.
- Built real-time geospatial driver search with sub-100 ms response times.
- Worked across event-driven workflows, service boundaries, order state, partner integrations and production migration.
FeedbackRig
May 2018 – Feb 2022 · Remote, United KingdomCTO
Led the technical direction and engineering of a multi-tenant customer-feedback SaaS platform.
- Built embeddable feedback, visual issue reporting, targeting, analytics and external integrations.
- Added NLP-based sentiment analysis across free-text feedback.
- Owned platform architecture, data processing and technical delivery.
Aspose
Apr 2011 – Dec 2021 · Remote, Sydney, AustraliaSoftware Architect
Built document-processing APIs and platform infrastructure handling 5M+ requests per month across 40+ file formats.
- Built identity and SSO infrastructure serving 1.6M users.
- Worked across API design, distributed processing, document conversion and platform modernization.
- Helped scale systems from early-stage traffic to sustained high-volume production workloads.
From building systemsTo sharing knowledge
Speaking
Talks and technical workshops
I speak and run technical sessions on enterprise AI, agent infrastructure, AI security, identity and distributed systems.
Talks
Selected talks
Workshops
Workshops for engineering teams
Practical sessions built around architectures, failure cases and implementation decisions from production AI and identity systems.
01
Building Agent Runtimes for Production
AI engineers · architects
Design the runtime around an agent rather than leaving orchestration inside the model loop.
- Context, tools and capability loading
- Durable state, retries and recovery
- Approval and execution flows
02
Attacking and Securing AI Agents
Security · AI · platform teams
Trace how untrusted context can become an authorized production action.
- Prompt injection and poisoned context
- Tool abuse and confused-deputy paths
- Execution controls and audit trails
03
Identity and Authorization for AI Agents
Architects · identity · security teams
Design how agents authenticate, receive authority and act on behalf of users or workloads.
- Agent and workload identity
- Delegation and task-scoped access
- Runtime authorization and revocation
04
Retrieval and Evaluation for Enterprise AI
AI engineers · technical teams
Build retrieval pipelines that can be measured instead of judged by whether an answer “looks right.”
- Query routing and hybrid retrieval
- Ranking, grounding and citations
- Evaluating unsupported answers
Speaking engagements
Discuss a Speaking Engagement Available for selected speaking engagements
Conference talks, corporate workshops and technical sessions for engineering, security and architecture teams.
Writing
Technical writing from systems I build and review
I write about agent architecture, secure execution, retrieval, identity, authorization and distributed systems, usually starting from a concrete architecture problem, implementation constraint or failure mode.
Featured article
Explore all writing Lost in Execution
How agent systems lose track of intent, authority and execution state as work moves through tools and services.
The Missing Runtime Between AI Agents and Enterprise Backends
Why production agents need a runtime layer for state, permissions, recovery and controlled execution.
Why Code Execution Is Eating Tool Registries
How code execution changes the way agents discover, compose and use capabilities.
The Future of Agents Is Outcome Coordination
A model for coordination based on objectives, commitments and evidence rather than shared orchestration.
Recommendations
From people I've worked with
When Hammad designs a system, he has already considered how it deploys, how the database evolves, how secrets and identities are handled, and what will fail the first time it runs in Kubernetes. … I would work with him again.
Nirav PatelDirector of Cloud OperationsView full recommendation
01 / 05
AdvisoryArchitecture and
Architecture and
technical leadership
I work with engineering, product and leadership teams on AI systems, identity platforms and enterprise software, from architecture reviews through implementation.
01
AI & Enterprise Architecture
Architecture for AI systems, identity platforms and enterprise software.
Focus
- System boundaries and service design
- Integrations and data flows
- Identity and authorization
- Runtime and security controls
02
Architecture & Security Review
Independent review of an existing system, proposed design or technical programme.
Focus
- Architecture assumptions
- Scalability and failure modes
- AI execution paths
- Identity and authorization
- Security controls
03
Embedded Technical Leadership
Continued architecture ownership through implementation and delivery.
Focus
- Technical direction
- Design reviews
- Engineering guidance
- Implementation decisions
- Production readiness
Contact
Start a conversation
For senior architecture roles, advisory work, speaking, workshops or research collaboration.
Choose a starting point or reach out directly.
01Senior architecture role
Architecture leadership, AI, identity or distributed systems.
Discuss a role 02Architecture / security reviewA system, design or technical problem you want another view on.
Start with context 03Speaking / workshopConference talk, team session or technical workshop.
Discuss an engagement 04Research / collaborationAgents, authorization, protocols or open-source work.
Explore collaboration