Platform & SRE
Production operations, incident response, observability, automation and operational product design.
Platform · SRE · Data · AI Infrastructure
Engineering systems that survive production.
Rishabh Gupta. Systems-minded engineering across platform, cloud, data and AI infrastructure, with a focus on evidence, architecture and the reality of operating complex systems.
10+years engineering
20+Spark / Delta workloads
~700Mrecord scenario
8case studies
AIengineering lab
PLATFORM
OBSERVABILITY
DATA
AI / RAG
AUTOMATION
SRE
SYSTEMS
BUILD → SCALE → OPERATE → EVOLVERG.001
01 / System map
Depth across the stack.
Grouped by engineering problems, not by logos. The landing page is the map. The portfolio is the evidence layer.
Data Platforms
Databricks, Spark, Delta, Kafka, database-aware write paths, data quality and throughput engineering.
AI Infrastructure
Local models, embeddings, vector search, RAG, MCP and agent workflows treated as infrastructure.
02 / Flagship work
Systems, not screenshots.
Each project exposes the problem, architecture, operating constraints and engineering trade-offs.
PRODUCTION / M3
M3 Operations Platform
Operational tooling connecting health monitoring, data search, workflows and production support across data, search and observability surfaces.
OPERATORS
/ USERS
/ USERS
PLATFORM
WORKFLOWS
WORKFLOWS
DATA / SEARCH
OBSERVABILITY
DATA / PERFORMANCE
Large-scale write-path engineering
A performance-engineering scenario around Databricks, Spark, staging and MSSQL delivery. The portfolio documents the reasoning around partitioning, batching, throughput, pressure and cost.
DELTA
SPARK
PARTITIONS
PARTITIONS
STAGING
MSSQL
ObserveLogs, metrics, traces and useful operational context.
RecoverReplay, idempotency, checkpoints and safe failure paths.
ControlClear boundaries, permissions and human control for risky actions.
EvolveModernization driven by evidence rather than novelty.
03 / Engineering lab
Experiments stay honest.
Research is clearly separated from production experience. The lab is about questions, observed behavior and next engineering moves.
AI / RETRIEVAL
RAG Stack Comparisons
Compare ingestion, chunking, embedding and retrieval behavior without hiding the pipeline behind model hype.
$ ingest → chunk → embed → Qdrant
$ retrieve → cite → evaluate → improve
$ lesson: pipeline quality > model switching alone
$ retrieve → cite → evaluate → improve
$ lesson: pipeline quality > model switching alone
ARCHITECTURE / AI
MCP & Agent Tooling
Tool-connected agents need guardrails, observability, data boundaries and environment discipline.
POLICY → TOOL ROUTER → DATA → ACTION
OBSERVE every call. Preserve provenance.
HUMAN controls high-risk actions.
OBSERVE every call. Preserve provenance.
HUMAN controls high-risk actions.
04 / Play
Engineering has a playground.
These are not portfolio decorations. They are interactive surfaces that show curiosity, visual craft and technical range.
05 / Operating model
How the work is approached.
Architecture means making trade-offs explicit. Reliability, cost, freshness, control and recovery are part of the design.
Design for failure.
Define failure paths before production discovers them.
Observe before optimizing.
Measure the bottleneck before changing the system.
Make data recoverable.
Partitioning, idempotency, replay and reconciliation matter.
Keep architecture proportional.
Use the simplest shape that meets the real constraint.