He practised the work. Now he builds the systems behind it.
Hassaan Mallick combines legal practice, teaching and engineering delivery to build governed AI workflows for law firms.
Before Hassaan Mallick designed AI systems, he spent more than a decade inside the work those systems now have to understand: contracts, compliance, due diligence, negotiation, disputes and training.
The legal work
His career began as a legal trainee at Standard Chartered Bank, working on contract drafting and review, regulatory compliance and pre-trial litigation support, with rotations through operations, trade, due diligence and documentation. He completed the bank's Anti-Bribery Risk Certification with full marks.
At Liaquat Merchant Associates — ranked by Chambers Global and Legal 500 — he worked on finance agreements, service contracts, client due diligence and negotiations. At GLOBALSERVE he handled legal work across five businesses, built the policies and systems behind the commercial process, and negotiated contracts with clients across 120 jurisdictions.
Practice and teaching
He then ran his own practice for four years across family, criminal and contract law. The work included white-collar crime, banking fraud and cybercrime support; company formation and investor funding; contract negotiation; and property disputes.
During the same period he lectured law at Ziauddin University, teaching legal skills, logic and reasoning, and public speaking to more than 200 students. He also spent three and a half years as a training and development consultant at WorkPlay Consulting.
Systems delivery
From 2023 to 2026 he ran delivery on web and application engineering projects for international clients at Finsweet. He now designs and builds production AI systems: architecture, orchestration and the controls that let a model operate inside a business process.
Shipped work includes the operational infrastructure behind an M&A advisory's outbound function — three systems covering intake, decisioning and reporting.
Hassaan no longer practises law. He completed the Legal Practice Course at the University of Law, Bloomsbury, but is not admitted as a solicitor in England and Wales and is not licensed in any market MallickTech serves. MallickTech provides technology services, not legal advice.
The model is only one part of the system.
Production work depends on the boundaries around it: permissions, state, recovery, traceability and review. Those controls are part of the architecture from the start.
Bounded judgement
Models interpret intent and select a path. The system constrains the prose, actions and approvals that can follow.
Tenant isolation
Classification and routing resolve from data. Credentials stay scoped per tenant; onboarding is configuration, not code.
State re-checked
Before a delayed second action fires, the external system of record is checked again. Stale state does not drive new work.
Recoverable failure
Every node has an error path. The originating payload is retained, race conditions have a defined recovery route and failed work can be replayed.
Complete trace
The event log records what arrived, how it was classified, where it went, when it moved and whether it succeeded.
Completion verified
Asynchronous vendor APIs are polled to completion rather than assumed to have finished.