Hello, I'm
Yarra Satwik

I'm a

Software Engineer

Backend & Data Systems Builder

Applied AI Developer

I build
Python & FastAPI Services • Distributed Data Systems • LLM & RAG Workflows • Retrieval Systems • Evaluation Harnesses • Cloud-Native Software •
Python & FastAPI Services • Distributed Data Systems • LLM & RAG Workflows • Retrieval Systems • Evaluation Harnesses • Cloud-Native Software •
My Skills

Languages

Backend & Data

AI & ML

Cloud & DevOps

...and many more

NowNowlet'slet'stalktalkaboutaboutmymyexperienceexperience

Junior Software Developer at Lalataksha Consulting Services Private Limited

  • Architected Python/FastAPI services and PostgreSQL document workflows with async job orchestration and Redis caching, cutting p95 API latency by 41% while supporting 15,000+ documents per month.
  • Built semantic chunking, embeddings, FAISS retrieval, schema-constrained outputs, and validation, raising held-out field extraction accuracy from 79% to 92%.
  • Added model/API regression gates and automated Jenkins delivery with health checks and rollback gates, reducing release effort by 60%.
PythonFastAPIPostgreSQLRedisJenkins

B.Tech in Computer Science Engineering at Vignan's Lara Institute of Technology & Science

Built a foundation in software engineering, data systems, machine learning, and production application development in Andhra Pradesh, India.

Computer ScienceBackend EngineeringApplied AI
My Projects

Trace

Institutional-memory runtime for software teams with GitHub webhook ingestion, transactional outbox delivery, tenant-scoped vector retrieval, schema-constrained Bedrock reasoning, and provenance-aware memory lifecycle in CockroachDB.

JANUS

Merchant-side authorization gateway for AI buyers combining signed mandates, deterministic hard-policy gates, evidence-bound semantic scoring, human step-up, and concurrency-safe execution.

Aspect-Based Sentiment Analysis

Ensemble NLP pipeline combining LSTM, GRU, BERT, and RoBERTa with aspect-aware preprocessing, achieving 92.1% macro-F1 and a 6.4-point gain over the strongest single-model baseline.

Phishing Detection

Ensemble classifier built from lexical, host, and URL-structure features across 50,000+ URLs, reaching 98.3% ROC-AUC with sub-50 ms local inference and calibrated thresholds.

...and many more

Achievements
Published peer-reviewed aspect-based sentiment analysis research in AIP Conference Proceedings.
Solved 200+ problems on CodeChef across multiple difficulty levels.
Thank you for visiting my portfolio!