SAHIL KHUNDIYA
Engineering distributed systems, intelligent retrieval pipelines, and production-grade AI architectures.Optimizing high-throughput backends and scalable AI infrastructure.
Observability
System Impact Metrics
Trajectory
Engineering Evolution
The progression from building interfaces to architecting high-scale intelligent infrastructure.
Monolithic Origins
Foundational Backend Engineering & REST APIs.
Distributed Transition
Scalable Microservices & System Integration.
Event-Driven Velocity
High-Throughput Asynchronous Pipelines.
Intelligent Retrieval
Production RAG & Vector Infrastructure.
Knowledge Orchestration
Graph-Augmented Intelligent Systems.
Elite AI Infrastructure
Architecting Scalable Hybrid AI Platforms.
Deep Dive
Hybrid AI Infrastructure
Engineering a high-fidelity retrieval system for production AI agents.
Optimizing Context Fusion
To move beyond simple RAG, we built a **Tri-Search Engine** that executes parallel queries across a **Knowledge Graph**, **Vector Index**, and **Full Text Search**. The key engineering challenge was "Context Fusion" — merging these disparate data sources into a ranked set of documents that minimize LLM hallucination while maximizing recall.
Custom intent classification layer using lightweight BERT models to route queries between exact search and semantic retrieval.
RRF (Reciprocal Rank Fusion) algorithm implementation to normalize scores from vector and keyword search pipelines.
Engineering Logic
Lower infrastructure overhead. By leveraging PG's native inverted indices alongside our vector store, we reduced network hop latency by 15ms.
Vector search captures "concepts", while FTS captures "specific identifiers". Combining them solved the 20% recall gap we faced with pure vector search.
Outcomes
Stack Overview
Evidence
Production Artifacts
Real-world system designs and performance optimizations.
Distributed Event Topology
Production-grade Kafka cluster architecture designed for 100k+ events/sec.
Query Plan Optimization
PostgreSQL P99 latency reduction through advanced indexing and query plan analysis.
Infrastructure
Live System Simulation
The lifecycle of an intelligent query through a distributed retrieval architecture.
Philosophy
Beyond the CRUD API.
My approach to engineering is centered on **reliability at scale**. I believe that any system that doesn't account for network failures, database lock contention, and traffic spikes isn't production-ready.
I prioritize **Clean Architecture** and **SOLID principles** not as dogmas, but as practical tools to reduce the cost of change in complex distributed systems.
"Simple is hard, but scalable is impossible without simplicity."
Fault Isolation
Implementing bulkhead patterns and circuit breakers to ensure service failures stay contained.
Event-Driven Systems
Leveraging asynchronous message brokers for decoupled architectures and reliable state propagation.
Observability-First
Designing systems with distributed tracing and structured logging for P99 latency monitoring.
Distributed State
Managing data consistency across microservices using Sagas and Outbox patterns.
Architecting for Resilience
Arsenal
Technology Stack
Java
LanguageSpring Boot
FrameworkKafka
StreamingPostgreSQL
DatabaseMicroservices
ArchitectureRAG Systems
AIKnowledge Graphs
Data StructureVector DB
AI InfrastructureDocker
DevOpsHibernate
ORMTrajectory
Engineering Experience
Amantya Technologies
Trainee Engineer – Software Development
Developing scalable backend microservices and intelligent systems.
- Engineered scalable backend microservices using Spring Boot and Hibernate/JPA.
- Implemented high-throughput async processing with Kafka and JMS.
- Optimized PostgreSQL queries for production-grade performance.
- Contributed to HLD/LLD for distributed system modules.
Samsung SDS
Developer Intern
Contributed to enterprise backend modules and system integration.
- Developed core backend modules using Java/Spring framework.
- Optimized complex SQL queries for data-intensive operations.
- Assisted in integrating various system modules into the main pipeline.
Coding Ninjas
Teaching Assistant
Mentored students in Data Structures and Algorithms.
- Solved over 1100+ DSA problems across various platforms.
- Assisted students in debugging complex algorithmic problems.
- Conducted doubt-clearing sessions for 500+ students.
Roadmap
Currently Exploring
The future of my engineering focus, moving towards deep infrastructure and autonomous agentic systems.
Distributed AI Systems
Scaling LLM inference and training across distributed clusters.
Graph Retrieval (GRAG)
Combining graph traversal with vector search for deep semantic context.
Reliability Engineering
Advanced observability and chaos engineering in backend systems.
Event-Driven AI
Real-time agentic workflows triggered by streaming data events.