# Knowledge, retrieval and context

> I give agents the context they need, from sources they can be held to.

RAG and GraphRAG, code knowledge graphs, retrieval with citations, structured extraction, and the data and context layers (metrics, repository standards) that agents read through MCP.

- Core skills: GraphRAG and RAG, Coverage and quality metrics, Model Context Protocol
- Domains: Telecom, Banking and financial services, Quality engineering, Legacy modernization
- URL: https://farhanbaluch.com/capabilities/knowledge-retrieval

## Case studies

- [Market research agents for retail banking](https://farhanbaluch.com/work/market-research-agents.md) (2025, Enterprise bank): 99.3%, extraction accuracy across 20+ bank websites. An LLM reads pages the way an analyst does, so the crawler survives layout changes that break scrapers. A vision model then checks every value against the rendered page before an analyst sees it.
- [Coverage intelligence, a quality data layer for agents](https://farhanbaluch.com/work/coverage-insights.md) (Oct 2026 – present, National telecom carrier): 81.7%, regression automation coverage across 21 teams, now measured weekly. Agents can only answer "how automated are we, and where should we invest?" if the numbers come from a source they can trust. This pipeline is that source: test data read straight from code, normalized across 21 teams, and structured to be served through an MCP server, so engineers and leadership can ask an agent for coverage instead of opening a spreadsheet. It is also how the AI test-automation programme shows whether test-generation agents are moving the number.
- [Agentic legacy modernization and code intelligence](https://farhanbaluch.com/work/legacy-modernization.md) (2025 – 2026, Enterprise bank): 70%, shorter modernization timeline. Spec-driven agents transform the code and a multimodal agent verifies migrated reports, while a GraphRAG platform gives engineers and agents a queryable map of the estate.
- [Estimation platform for an enterprise bank](https://farhanbaluch.com/work/estimation-platform.md) (2025, Enterprise bank): 80%, less manual estimation effort. Agents research and draft the estimate, a regression model anchors the number, retrieval over past projects supplies citations, and an expert approves. The AI does the slow part and a person keeps the decision.
- [GenAI solution architecture and pre-sales](https://farhanbaluch.com/work/genai-solution-architecture.md) (Oct 2024 – 2025, Enterprise banking and telecom clients): 3, enterprise client wins, from 10+ proofs of concept. Most enterprise AI work stalls between "this is interesting" and "this is approved". Reusable blueprints for RAG, agents and guardrails let each proof of concept start from a sound architecture, and the briefings turned leadership questions into pilots with a clear measure of success.
- [Multi-agent legacy documentation](https://farhanbaluch.com/work/legacy-docs-agents.md) (Aug 2026 – Oct 2026, Client demo, embedded systems): 16 sections, every claim verified against cited lines. Modernizing with AI starts with understanding the old system, and an LLM summary that hallucinates is worse than none. Deterministic analysis, explorer agents and an independent reviewer together produce documentation people can trust, and that becomes context for the agents that do the migration.

## Evidence from the log

- 2026-10-19: **Extended coverage reporting to mobile test suites.** Mobile teams write tests in Java with TestNG, which the first parser couldn't read. Added a Java-aware parser so their coverage shows up in the same weekly report. AI angle: An agent asked "how covered is the app?" needs every platform in one model, or it answers for web only. This closes the gap so the coverage data an MCP server will serve to agents describes the whole product. [Knowledge, retrieval and context; AI adoption at enterprise scale] (https://farhanbaluch.com/work/coverage-insights.md)
- 2026-10-14: **Published a structure and naming standard for 20+ test repositories.** Every team laid out its repository differently, which made automated reporting brittle. Wrote one standard for folders, tags and naming, with a migration checklist per team. AI angle: A predictable layout is the context layer agents need. With one standard, every agent (test generation, review, coverage queries) finds tests, page objects and tags in the same place in every repository, instead of each agent being tuned per team. [Knowledge, retrieval and context; AI adoption at enterprise scale] (https://farhanbaluch.com/work/coverage-insights.md)
- 2026-10-06: **Shipped the first weekly automation-coverage report.** 21 teams, measured straight from code instead of spreadsheets. The first run put regression automation coverage at 81.7%. AI angle: Gives the AI test-automation programme a baseline measured from code, so the effect of test-generation agents shows up in the numbers, and gives future agents a trusted source to query. [AI adoption at enterprise scale; Knowledge, retrieval and context] (https://farhanbaluch.com/work/coverage-insights.md)
- 2026-09-29: **Started coverage measurement from source.** Leadership's automation numbers came from hand-kept spreadsheets nobody could trace. Prototyped a zero-dependency pipeline that clones every team repository and counts tests by tag. AI angle: The first step toward a quality data layer agents can read through MCP. An agent can't reason about coverage when the only source is a spreadsheet. [Knowledge, retrieval and context; AI adoption at enterprise scale] (https://farhanbaluch.com/work/coverage-insights.md)
- 2026-08-18: **Started the legacy documentation agents.** Modernization projects stall because nobody understands the old code. Began an agent pipeline that combines deterministic static analysis with parallel LLM explorers. AI angle: Deterministic tools handle what must be exact (call graphs, registers) and agents handle what needs reading and explanation. The result becomes context for the agents that do the migration. [Agent architecture and orchestration; Knowledge, retrieval and context] (https://farhanbaluch.com/work/legacy-docs-agents.md)
- 2026-01-19: **Mapped a legacy codebase into a knowledge graph.** Turned repositories into a Neo4j graph of modules, calls and data flows, so engineers and agents query structure instead of grepping. Code navigation got 70% faster. AI angle: GraphRAG over code. Agents get relationships (who calls what, where data flows) that plain vector search misses. [Knowledge, retrieval and context; Agent architecture and orchestration] (https://farhanbaluch.com/work/legacy-modernization.md)
- 2025-07-07: **Grounded estimates in historical projects.** Added retrieval over past estimates with Azure AI Search and a regression baseline on Databricks, so every number comes with a confidence interval and citations. AI angle: Grounding keeps the LLM from inventing numbers. It adjusts a statistical baseline and cites the past projects it used. [Knowledge, retrieval and context; Applied ML and research] (https://farhanbaluch.com/work/estimation-platform.md)
- 2025-04-14: **Reached 99.3% accuracy on bank-site extraction.** Added a validation pass that screenshots each page and has a vision model confirm every extracted value. Two weeks of market analysis now runs in under an hour. AI angle: One model extracts and another, with a different input (the screenshot), verifies. Cross-modal checking is what moved accuracy into the range analysts trust. [AI evaluation and quality; Knowledge, retrieval and context] (https://farhanbaluch.com/work/market-research-agents.md)
- 2025-03-03: **Built an LLM crawler for 20+ bank websites.** Selector-based scrapers broke every time a site changed. The crawler lets a model read rendered pages into a fixed schema instead. AI angle: Uses the LLM for what it's good at, reading messy pages like a person, and constrains it with a fixed output schema. [Knowledge, retrieval and context; Agent architecture and orchestration] (https://farhanbaluch.com/work/market-research-agents.md)
- 2025-01-27: **Drew Azure and AWS architectures for GenAI proofs of concept.** Reference designs for RAG, agent hosting and guardrails that pre-sales could adapt per client. They backed 10+ proofs of concept and 3 client wins. AI angle: Reusable blueprints for RAG, agents and guardrails, so each proof of concept starts from a sound architecture instead of a blank page. [Solution architecture and delivery; Knowledge, retrieval and context] (https://farhanbaluch.com/work/genai-solution-architecture.md)
- 2024-10-21: **Joined mobileLIVE (now ML Arteka) as an AI Engineer.** First weeks on proofs of concept for enterprise banking clients, from RAG assistants to agent workflows. AI angle: Moved from research into building LLM systems for clients, starting with RAG and agent proofs of concept. [Solution architecture and delivery; Knowledge, retrieval and context] (https://farhanbaluch.com/work/genai-solution-architecture.md)
- 2023-05-15: **Built crop-disease detection and a policy chatbot.** At Agriculture and Agri-Food Canada, built computer vision for early crop-disease detection and a chatbot that answers farmers' policy questions. AI angle: Vision models and a retrieval chatbot over policy documents. My first RAG system, in a public-sector setting. [Applied ML and research; Knowledge, retrieval and context] (https://farhanbaluch.com/work/applied-ai-prototypes.md)