LogicLens
Fractional data architecture and AI consulting for teams that need senior-level design and delivery — without the overhead of a full-time hire.
Get in touchI embed as your senior data architect on a part-time basis — designing systems, reviewing stack decisions, and setting standards your team can execute against. You get the strategic clarity without the cost of a full-time principal hire.
I design and build production-grade retrieval-augmented generation pipelines end to end — vector store selection, indexing strategy, chunking, re-ranking, and API integration. No prototypes handed over the fence; I stay through deployment.
Migrating off legacy infrastructure to Snowflake, Databricks, or Delta Lake? I've run these projects at scale — including a Rackspace-to-Snowflake migration that delivered $900K in annual savings. I scope, architect, and lead delivery.
Aerospace & Defense
Built a production RAG pipeline on Azure GovCloud for a large aerospace manufacturer, handling technical documentation retrieval in an ITAR-cleared environment. Architecture: Databricks medallion lakehouse, LanceDB with IVF_HNSW_SQ indexing, and a retrieval layer tuned for precision over noisy unstructured text. Deployed fully within GovCloud compliance boundaries.
Retail & Marketing
Modernized a large retailer's external marketing data pipelines — migrating fragile legacy processes onto a Snowflake-based platform with dbt for transformation and Python for orchestration. The migration off the old hosting platform delivered $75K/month in savings, $900K annually. The marketing team went from unreliable batch jobs to pipelines they could actually trust.
Decision makers don't hire architects when the path is clear. They hire them when it isn't. Here's the framework I use to find clarity fast — every engagement, every domain.
Read more →I spent a Saturday afternoon building a production-grade RAG pipeline on 300,000 records. A senior dev fluent in the full stack would have needed two to three weeks. Here's what actually made that possible — and what it means for how I deliver for clients.
Read more →AI is on every roadmap. Few organizations have stress-tested whether their data foundation can actually support it. Here's the pattern I keep seeing — and how I assess it.
Read more →I've built enterprise AI chat twice. Once clean. Once in the real world. The gap between the two is where most initiatives get into trouble — and where the real work lives.
Read more →Most teams instrument their AI systems for system health. Few capture the signals that actually matter. Here's what your metrics database is probably missing — and why it compounds.
Read more →In Inception the most dangerous idea is the one planted so deep it becomes your own original thought. That's what happened to me with CLAUDE.md. This is that story.
Read more →Every team owns their layer. Nobody owns the intersections. A five-part series on data pipelines, designation, and the instrumentation that ends the midnight call.
Read the series →Auto Loader. Structured Streaming. LanceDB. The tools work. The complexity lives at the seams. A six-part series on building near real time vector search in production.
Read the series →Every enterprise document project hits the same wall — a Teams thread about column size. That's not a column size problem. It's an architecture problem. Here's the fix.
Read more →RAG is a retrieval problem. Production is an observability problem. A six-post, opinionated, practitioner-built blueprint for what every AI system actually needs before it earns the word production.
Read the series →I'm Arjun Krishnamoorthi, a data architect and engineer based in Chicago with 25+ years working across the full data stack — from pipeline infrastructure to analytical systems to production AI. I've worked with Fortune 500 enterprises and early-stage startups, and I've led projects ranging from greenfield platform builds to complex legacy migrations.
Some of that work has been in regulated defense environments — I've held ITAR clearance and understand what it means to design systems where compliance is non-negotiable.
I run LogicLens LLC as a fractional practice. My clients typically need a senior architect who can think through a problem end-to-end, make defensible decisions quickly, and work alongside their existing team without hand-holding or overhead. That's what I do.
I build on the stack that's right for the engagement. For LogicLens demos and greenfield projects that's Anthropic's Claude — available across AWS, Azure, and GCP. For client engagements I work within your approved stack, including regulated and government environments. The architecture is sound regardless of the stack. That's the point.
If you have a data architecture problem or an AI project that needs senior hands — let's talk.