Architecting the intersection of Information Retrieval and Generative AI.
With over a decade of experience in Search (Solr/Elasticsearch/OpenSearch), I have pivoted to engineering production-grade GenAI systems. For the last 5 years, I specialize in building Retrieval-Augmented Generation (RAG) platforms, scaling LLM agents, and optimizing vector search for enterprise environments.
My focus is moving beyond "prompt engineering" to build robust, secure, and observable AI architectures that solve complex data access problems.
I am currently working on GraphRAG, Local LLM inference, and AI Agents.
🏴 ctf-kit: An offensive security agent framework that integrates with Claude Code and Copilot. It doesn't just analyze code; it orchestrates collaborative reasoning to detect binary vulnerabilities and synthesize exploit scripts for Capture The Flag challenges in real-time.
🧠 Adaptive Knowledge Graph: A neuro-symbolic learning engine running entirely on consumer hardware. Fuses structured Knowledge Graphs with LLM reasoning to simulate an AI tutor that adapts to student cognitive states using Bayesian Knowledge Tracing—zero cloud dependency required.
🧱 AgentBricks: Architectural primitives for deploying autonomous agents on the Data Lakehouse. Enables LLMs to reason directly over Unity Catalog volumes, turning static enterprise data into active, queryable knowledge assets without data movement.
🎙️ whisper-danger-zone: An air-gapped audio intelligence pipeline. Orchestrates state-of-the-art Whisper models with Pyannote diarization to transmute raw audio into speaker-attributed transcripts, ensuring 100% data privacy for sensitive signal processing.
| Domain | Stack & Tools |
|---|---|
| GenAI & LLM | Amazon Bedrock, Azure OpenAI, LangChain, RAG Architectures, Local LLMs (Ollama/Llama.cpp), Prompt Security (OWASP) |
| Search & Data | Elasticsearch, OpenSearch, Solr, Lucene, Vector Databases, Hybrid Search (Lexical + Semantic) |
| Engineering | Python (Deep Ecosystem), Java, AWS, Databricks, API Design, System Architecture |
| Niche | Molecule Similarity (Cheminformatics), Browser Fingerprinting, NLP/NER (Spacy, Flair) |
Principal Engineer | Enterprise GenAI Platform
- RAG at Scale: Architected a central RAG API acting as a proxy between internal engineering hubs and Amazon Bedrock. The system aggregates knowledge from tech documentation, metadata, and tooling catalogs to power a developer-focused assistant.
- LLM Gateway: Led the technical strategy for abstracting model providers, allowing teams to switch between models while maintaining consistent security and observability standards.
GenAI Architect | Blueprint & Security
- Architecture Strategy: Spearheaded the "GenAI Blueprint," a reusable architecture used to bootstrap multiple internal applications, including a QnA chatbot and review summarization tools.
- Security First: Implemented strict adherence to OWASP Top 10 for LLMs to mitigate prompt injection and data leakage risks in a corporate environment.
Search Optimization
- High-Volume Retrieval: Optimized search and recommendation engines for a leading e-commerce platform, focusing on both "quick win" relevancy tuning and long-term hybrid search transformations to improve customer purchase journeys.
I love sharing knowledge about the transition from traditional search to modern AI-driven retrieval.
- Python Generators for Search Engines (Summer Python Meetup) - Watch (RU) | Slides
- Deploying Solr in Multi-Region Environments (Apache Lucene/Solr London) - Event Link
- Effective Molecule Search in Elasticsearch (Cambridge Cheminformatics & Zed Conf) - Watch | Slides
- Browser Fingerprinting & Privacy - Slides
- CTF Competitions (Codeberry Club) - Watch
