Aviral Harsh
Areas of Expertise
- Systems Architecture & Microservices
- High-Concurrency Backend Engineering
- Big Data Processing & ETL Optimisation
- Event-Driven Design
- Caching Strategies
- ML/AI Model Deployment
- API Development & Performance Tuning
- Production Monitoring & Observability
- Cloud Infrastructure & Cost Optimisation TECHNICAL Proficiencies Frontend Technologies: React, React Native, Redux, Next.js | TypeScript | Material-UI, Tailwind CSS Programming Languages: Java, Python, Scala, SQL, JavaScript / TypeScript, C++ AI / Machine Learning: PyTorch, TensorFlow | Chunking, Embeddings, Reranking | LangChain, RAG | Hugging Face, Fine-Tuning Cloud & DevOps: AWS (EC2, S3, Lambda, EKS, RDS, SQS, SNS, Athena) | Docker, Kubernetes, Helm | Terraform | CI/CD Pipelines Data Engineering: Apache Spark, Airflow, Druid | Hadoop, Hive | Presto, Trino | PostgreSQL | MongoDB, Redis Systems: Spring Boot, Spring Cloud, Microservices, Event-Driven Architecture, REST APIs, GraphQL, gRPC | Node.js, FastAPI | RabbitMQ
Experience
Software Development Engineer-1
Nielsen
Remote, Gurgaon, India
- Delivered 99.95% platform availability and sub-250ms p99 latency by designing an event-driven microservices architecture with Java, RabbitMQ, Spring Boot, Spring Cloud, and Resilience4j circuit breakers to isolate faults and prevent cascading failures.
- Reduced cascading failures by 70% and decreased payload sizes by 40% by modernising internal synchronous service communication to gRPC and implementing resilience patterns across distributed services.
- Enabled reliable handling of 100K requests per day with ~200ms response times by optimising API performance through Redis caching, asynchronous workflows.
- Accelerated ETL runtimes by 42% across 8TB+ daily data pipelines by optimising Spark joins, indexing, correcting partitioning strategies, and mitigating data skew via salting techniques.
- Maintained 99.7% orchestration success rate across 280+ Airflow DAGs by strengthening workflow dependency management, retry logic, and failure recovery mechanisms.
- Improved Druid query performance by 2.8× and lowered cross-system latency by 35% by tuning segment granularity, enabling bitmap indexing, and implementing Trino federated queries across PostgreSQL, S3, and Druid.
Software Development Engineer Intern
Rigi
Bengaluru, India
- Generated $28K+ ARR and supported 81K+ active users by delivering 17 creator applications with OAuth authentication and payment integrations across React, React Native, Java, Spring Boot, and PostgreSQL stack.
- Achieved 77% fraud-detection precision with ~410ms inference latency at 11K requests/sec by building TensorFlow and XGBoost ensemble models incorporating LSTM autoencoders, deployed on Kubernetes.
- Increased analytics throughput by 6× whilst processing 8M+ events/day by architecting a Kafka and ClickHouse-driven streaming pipeline enabling sub-second query performance.
Software Development Engineer Intern
Amazon
Chennai, India
- Reduced p95 latency from 420ms to 165ms and lowered API calls by 35% by implementing GraphQL query batching, response caching, and backend performance tuning.
- Sustained 99.98% legitimate throughput under high-traffic conditions by engineering a Redis-backed distributed token- bucket rate-limiting solution deployed on AWS Lambda.
- Achieved 92% automated test coverage and decreased production defects by 40% by using JUnit and Mockito.
Education
Birla Institute of Technology and Science (BITS Pilani)
B.E.(Hons) in Electrical and Electronics Engineering | Master of Science in Biological Sciences (Dual Degree)
Hyderabad, India | GPA: 7.9 / 10
Delhi Public School (CBSE) – High School DPS Bokaro
India | Class 12: 95% | Class 10: 10 GPA
Skills
KEY PROJECTS
AI Study Assistant
Achieved 50× document retrieval acceleration and 90% answer accuracy by building LangChain + FAISS + Ollama RAG pipeline processing 5K+ page PDFs with semantic chunking and vector caching; integrated live web search.
AI Resume Builder
Engineered a FastAPI platform using vector embeddings and LLM prompt chaining for automated, sub- 4s resume editing and cover letter synthesis.
Conversational Voice AI Agent (STT + LLM + TTS)
Engineered a sub-1s TTFA voice agent by building an asynchronous text-to-audio streaming architecture (FastAPI, WebSockets). Integrated Deepgram/Whisper, Groq, and ElevenLabs, optimizing conversational fluidity via semantic LLM token chunking and efficient context memory state management.