AI Engineering Portfolio
RAG, Document Intelligence & LLM Workflows
A selection of AI engineering projects spanning knowledge retrieval, contract analysis, dashboard assistants and reliable backend workflows.
Engineering Focus
These projects explore how language models connect to real application workflows: retrieving access-controlled knowledge, extracting structured contract data, selecting dashboard context and coordinating asynchronous work. The emphasis is on explicit data contracts, bounded inputs and outputs, streaming interfaces and testable backend services.
- Hybrid retrieval & graph-enhanced RAG
- Schema-validated document intelligence
- LangGraph routing & streaming APIs
- Durable queues, leases & retry boundaries
Technical Skills
Tools and engineering practices represented in the projects below, from retrieval and document processing to streaming APIs and asynchronous work.
LLM Workflows
- LangGraph
- OpenAI Responses API
- Structured Outputs
- Prompt Routing
- Context Budgets
Knowledge & Retrieval
- Hybrid RAG
- Neo4j
- Embeddings
- Rank Fusion
- Reranking
- Access-Aware Retrieval
Document Intelligence
- Structured Extraction
- Contract Summaries
- Contract Alerts
- Document Comparison
- PDF Processing
Backend & Interfaces
- Python
- FastAPI
- Pydantic
- PostgreSQL
- SQLAlchemy
- Next.js
- SSE
Asynchronous Systems
- Celery
- RabbitMQ
- Redis
- Idempotency
- Work Leases
- Docker
- pytest
Featured Projects
Five projects covering knowledge retrieval, document intelligence, dashboard assistants and backend workflow infrastructure.
Knowledge Wiki AI
Access-Aware Knowledge Retrieval
A knowledge platform that answers questions using access-aware, graph-enhanced hybrid retrieval with streamed responses.
- Semantic and full-text search combined with Reciprocal Rank Fusion
- Graph expansion, reranking and a confidence gate before generation
- Version-aware asynchronous ingestion, chunk-level access filters and SSE sources
Document Control AI
Structured Contract Intelligence
A document-processing API for structured extraction, contract summaries, configurable alerts and document comparison.
- Schema-validated LLM extraction and structured contract summaries
- Contract assessments combined with extracted deadlines and timing information
- PDF, DOCX, JSON and image processing with bounded downloads and host allowlists
Dashboard Query Assistant
Context-Bounded Analytics Chatbot
A dashboard assistant that routes questions and prepares relevant data for LLM-powered analytics responses.
- LangGraph orchestration for general questions and dashboard query flows
- Relevant dashboard rows selected and packed within token and byte budgets
- Synchronous and SSE query endpoints with rate limiting and request observability
Kworks AI
Prompt Orchestration Service
Phase 1 · Static knowledgebase
A prompt-orchestration service for knowledge, content, SEO and classification workflows using approved static context. Retrieval and ingestion remain future phases.
- LangGraph routes requests to task-specific prompts with shared policies
- Allowlisted knowledge packs with source hashes and prompt version tracking
- Structured Outputs, source and label validation, streaming and request deadlines
Reporter AI
GitLab Review Infrastructure
In development · Infrastructure foundation
Infrastructure for a GitLab review pipeline, with durable webhook intake and bounded processing primitives. The end-to-end reviewer and comment publisher are not yet implemented.
- Authenticated System Hooks persisted to a deduplicated PostgreSQL inbox before HTTP 202
- Bounded dispatch batches, fenced work leases and shared retry budgets
- Offline diff parsing, prepared model input and evidence-anchor validation