Enterprise Knowledge Intelligence & Grounded RAG:
Weekly 3-Hour Weekend Masterclasses.
Experience our high-bar engineering pedagogy live before committing to the full 12-week foundry. Every weekend, Founder & Principal AI Architect Soham Sharma leads an intensive 3-hour code-along building production slices of the 6-layer Enterprise Knowledge Intelligence Platform.
Zero Financial Risk Guarantee
100% of your ₹999 / $19 masterclass fee is credited as a direct tuition deduction toward the March 2027 Forward Deployed AI Engineer cohort.
Time & Format
Saturday or Sunday nights · 09:00 PM – 12:00 AM IST (11:30 AM – 02:30 PM EST). 100% live interactive coding over Zoom.
Runnable Deliverables
Walk away with complete, documented GitHub starter repositories, Docker configurations, and test suites ready for production deployment.
Direct Founder Access
Taught personally by Soham Sharma. Includes a 30-minute open architecture review session to inspect your own system designs.
Upcoming Weekend Masterclasses
Each 3-hour session tackles one specific production bottleneck in enterprise AI systems engineering.
Building an Evidence-Grounded Abstention Engine in LangGraph
Move beyond hallucinated answers. Learn how to extract atomic factual claims from LLM completions, verify span-level source citations, and implement a deterministic abstention node in LangGraph that refuses to guess when retrieved evidence is absent or contradictory.
Verified Code Deliverable:
Runnable LangGraph Abstention Engine repository with DeBERTa entailment checking.
Ingesting Multi-Column PDFs & Borderless Tables with Docling
Dissect why traditional PDF text converters silently corrupt downstream RAG pipelines. Code a production layout-aware extraction engine using Docling and Unstructured, reconstructing complex financial tables and attaching cryptographic SHA-256 provenance hashes.
Verified Code Deliverable:
Docling Table & Blueprint Extraction Pipeline with bounding box provenance.
Neo4j GraphRAG: Triples, Ontologies & Multi-Hop Cypher Search
When vector search fails on relational queries. Learn to model domain ontologies, extract subject-predicate-object triples with structured extraction guards, and execute multi-hop Cypher traversals in Neo4j for deep enterprise reasoning.
Verified Code Deliverable:
Production Neo4j Knowledge Graph with live Cypher query interface.
Hybrid RRF Search & Production Observability with Prometheus & Grafana
Combine sparse BM25 with dense pgvector vectors using Reciprocal Rank Fusion (k=60), apply Cohere Rerank 5 Pro, and build Prometheus and Grafana dashboards tracking TTFT, token cost, and Ragas faithfulness metrics.
Verified Code Deliverable:
Hybrid Retrieval Engine with Docker Compose Prometheus/Grafana stack.
Reserve Your Weekend Masterclass Seat
Seats are capped at 50 engineers per session to ensure live code-along interactivity. 100% of the ₹999 INR / $19 USD fee is credited toward the March 2027 cohort.