Software Engineering
DefendaMe
AI-assisted legal discovery for evidence-heavy casework, combining document processing, search, transcription, background workflows, traceability, and reliable AI analysis.
Year :
2025–2026
Area :
LegalTech / Digital Evidence / AI
Organization :
DefendaMe
Project Duration :
Ongoing
The problem
Legal teams work with large amounts of documents, recordings, digital evidence, deadlines, and repeated analysis. The hard part is not simply adding AI. It is making search, automation, and AI useful inside real casework without losing reliability or traceability.
DefendaMe was built around that constraint: help legal professionals work through complex case material while keeping the system understandable, auditable, and operationally reliable.
What I built
As Technical Founder, I work across backend, product infrastructure, and the workflows that connect evidence to the user experience.
I built APIs and background workers for document ingestion, OCR, transcription, search, case workflows, and AI-assisted analysis. The platform grew to six BullMQ queues and fourteen workers, with more than 833,000 indexed pages, 260,000 OCR-processed pages, and about 31 GB of files processed across real workflows.
I also worked directly with legal and forensic specialists to understand how the software needed to behave in actual casework, then translated those requirements into product and architecture decisions.
Engineering challenges
Most of the engineering work came from making many moving parts behave as one system: long-running jobs, search indexing, transcription, AI providers, user-facing progress, retries, costs, and failure recovery.
I added idempotency, dead-letter handling, rate limits, backpressure, correlation IDs, structured logging, and operational metrics so failures could be understood and recovered instead of becoming hidden state.
Evidence handling added another constraint: actions needed to remain traceable. I worked on append-only chain-of-custody events, structured audit data, and redaction of sensitive information alongside the rest of the application infrastructure.
What I learned
DefendaMe gave me practical experience across product discovery, backend architecture, asynchronous systems, search, AI infrastructure, observability, reliability, and sensitive-data workflows.
One of the main lessons was that operational concerns such as failures, retries, costs, recovery, and traceability need to be considered as part of the product itself, especially in workflows where users depend on the system for high-stakes work.
More Projects
Software Engineering
DefendaMe
AI-assisted legal discovery for evidence-heavy casework, combining document processing, search, transcription, background workflows, traceability, and reliable AI analysis.
Year :
2025–2026
Area :
LegalTech / Digital Evidence / AI
Organization :
DefendaMe
Project Duration :
Ongoing
The problem
Legal teams work with large amounts of documents, recordings, digital evidence, deadlines, and repeated analysis. The hard part is not simply adding AI. It is making search, automation, and AI useful inside real casework without losing reliability or traceability.
DefendaMe was built around that constraint: help legal professionals work through complex case material while keeping the system understandable, auditable, and operationally reliable.
What I built
As Technical Founder, I work across backend, product infrastructure, and the workflows that connect evidence to the user experience.
I built APIs and background workers for document ingestion, OCR, transcription, search, case workflows, and AI-assisted analysis. The platform grew to six BullMQ queues and fourteen workers, with more than 833,000 indexed pages, 260,000 OCR-processed pages, and about 31 GB of files processed across real workflows.
I also worked directly with legal and forensic specialists to understand how the software needed to behave in actual casework, then translated those requirements into product and architecture decisions.
Engineering challenges
Most of the engineering work came from making many moving parts behave as one system: long-running jobs, search indexing, transcription, AI providers, user-facing progress, retries, costs, and failure recovery.
I added idempotency, dead-letter handling, rate limits, backpressure, correlation IDs, structured logging, and operational metrics so failures could be understood and recovered instead of becoming hidden state.
Evidence handling added another constraint: actions needed to remain traceable. I worked on append-only chain-of-custody events, structured audit data, and redaction of sensitive information alongside the rest of the application infrastructure.
What I learned
DefendaMe gave me practical experience across product discovery, backend architecture, asynchronous systems, search, AI infrastructure, observability, reliability, and sensitive-data workflows.
One of the main lessons was that operational concerns such as failures, retries, costs, recovery, and traceability need to be considered as part of the product itself, especially in workflows where users depend on the system for high-stakes work.
More Projects
Software Engineering
DefendaMe
AI-assisted legal discovery for evidence-heavy casework, combining document processing, search, transcription, background workflows, traceability, and reliable AI analysis.
Year :
2025–2026
Area :
LegalTech / Digital Evidence / AI
Organization :
DefendaMe
Project Duration :
Ongoing
The problem
Legal teams work with large amounts of documents, recordings, digital evidence, deadlines, and repeated analysis. The hard part is not simply adding AI. It is making search, automation, and AI useful inside real casework without losing reliability or traceability.
DefendaMe was built around that constraint: help legal professionals work through complex case material while keeping the system understandable, auditable, and operationally reliable.
What I built
As Technical Founder, I work across backend, product infrastructure, and the workflows that connect evidence to the user experience.
I built APIs and background workers for document ingestion, OCR, transcription, search, case workflows, and AI-assisted analysis. The platform grew to six BullMQ queues and fourteen workers, with more than 833,000 indexed pages, 260,000 OCR-processed pages, and about 31 GB of files processed across real workflows.
I also worked directly with legal and forensic specialists to understand how the software needed to behave in actual casework, then translated those requirements into product and architecture decisions.
Engineering challenges
Most of the engineering work came from making many moving parts behave as one system: long-running jobs, search indexing, transcription, AI providers, user-facing progress, retries, costs, and failure recovery.
I added idempotency, dead-letter handling, rate limits, backpressure, correlation IDs, structured logging, and operational metrics so failures could be understood and recovered instead of becoming hidden state.
Evidence handling added another constraint: actions needed to remain traceable. I worked on append-only chain-of-custody events, structured audit data, and redaction of sensitive information alongside the rest of the application infrastructure.
What I learned
DefendaMe gave me practical experience across product discovery, backend architecture, asynchronous systems, search, AI infrastructure, observability, reliability, and sensitive-data workflows.
One of the main lessons was that operational concerns such as failures, retries, costs, recovery, and traceability need to be considered as part of the product itself, especially in workflows where users depend on the system for high-stakes work.