Vulnerability Intelligence Platform
Turn fragmented vulnerability data into an actionable remediation workflow.
ResultQualitative: shipped product capabilities across the remediation workflow and its supporting platform.
- Role
- Full-stack engineering within a cross-functional team
- Period
- 2025 - 2026
- Context
- Vulnerability intelligence
- Confidentiality
- Anonymized professional work
At a glance
Turn fragmented vulnerability data into an actionable remediation workflow.
Built product-facing remediation features across React and Go, integrated Bedrock-backed LLM output, shaped PostgreSQL data flows, and supported AWS delivery.
Qualitative: shipped product capabilities across the remediation workflow and its supporting platform.
The choice behind the implementation.
- Challenge
- Model-assisted remediation needed to be useful without asking engineers to trust an opaque answer.
- Decision
- Treat generated guidance as an inspectable artifact with source context, workflow state, and an explicit human review boundary.
- Trade-off
- The workflow carries more visible state, but engineers retain the evidence required to make the final decision.
The problem
Security teams need to move from vulnerability signals to a remediation decision without losing context across repositories, issue trackers, artifacts, and deployment systems. The product had to make that workflow usable while coordinating several technical boundaries.
Contribution boundary
The work crossed the React product surface, Go backend services, PostgreSQL data flows, model-assisted remediation, and AWS delivery infrastructure. Product and technical decisions were made with the wider team; this was not a solo project.
- 01Signal intake
Normalize vulnerability context from connected systems.
- 02Evidence model
Connect affected assets, source context, and workflow state.
- 03Assisted reasoning
Generate a proposed remediation with bounded context.
- 04Human review
Expose the proposal and supporting evidence for a decision.
- 05Delivery path
Coordinate approved changes with connected engineering tools.
- 06Feedback
Capture outcomes that can improve later recommendations.
Decision 01: make model output inspectable
AI-generated remediation is useful only when engineers can understand the source context, proposed change, and workflow status. The interface therefore treated the model response as a reviewable artifact, not an unquestionable answer.
Decision 02: treat integrations as system boundaries
Repository, issue-tracker, and artifact integrations fail differently. The backend work kept those boundaries explicit so retries, partial failures, and state transitions could be reasoned about without collapsing the workflow into one opaque request.
Decision 03: build delivery into the feature
The platform work included container and infrastructure concerns through ECS, ECR, CloudFormation, CodePipeline, and GitHub Actions. Shipping the capability and observing its runtime behavior were part of the same engineering responsibility.
The public boundary is deliberate.
Product and client identities, customer data, repositories, and confidential implementation details are intentionally omitted. The page describes only confirmed personal contribution and safe technical context.