Skip to content
Shivam Gupta
Architecture Deep Dives

Case Studies

The design decisions behind the systems that run unattended — what each one had to get right, and what it replaced.

Retrieval & Agents

RAG Support Agent

Retrieval-grounded ticket resolution over the brochure corpus, with structured output and an explicit escalation path to a human for refunds, complaints and account issues.

800/80
Chunk / overlap
Top-8
Retrieval depth
2-way
Reply or escalate
Role

Designed the retrieval pipeline, output contract and escalation logic

The brochure corpus is chunked at 800 tokens with 80 overlap into a Qdrant vector store using Gemini embeddings, retrieved top-8. The model is constrained to answer only from retrieved passages and emits a structured object carrying both the reply and an escalation flag. The constraint is the product — a confidently wrong reply about a refund costs more than no reply at all.

Read case study
Operational Alerting

Daily Sales-Performance Alerting

Joins counsellor rosters, walk-in logs and live MCube telephony data, computes role-specific call targets and escalates below-target agents to their managers automatically.

3
Role-specific targets
20:00 IST
Daily run
13
Branches covered
Role

Built the data joins, target arithmetic and escalation mail

Targets differ by role — counsellor, team lead and telecounsellor — and are adjusted down per walk-in handled. Only agents below target are emailed, with their manager copied. What used to be a manager reading a spreadsheet each night is now an exception report that arrives on its own.

Read case study