About
Software Developer with ~1 year of experience building production REST APIs and backend services in FastAPI and PostgreSQL, with AI integration work spanning RAG pipelines, AWS Bedrock, and n8n workflows. At Avasoft, I owned the ICP data-management module for client Zywave on a FastAPI + PostgreSQL stack, and authored the end-to-end AWS solution design for an AI-powered QA automation platform for client Liberty Home Guard — scaling review from manual sampling to full-coverage automated scoring. Comfortable owning features end-to-end, writing clean, well-tested (Pytest) code, and collaborating with cross-functional teams.
Technical Skills
Languages
PythonSQLJavaScriptTypeScript
Backend & APIs
FastAPIREST API DesignPydanticORM
Databases
PostgreSQLpgvector
AI
RAG Pipelinesn8n Workflow
Testing & Tools
PytestPostmanDockerGitHubExpo
Frontend & Mobile
ReactReact Native
Experience
Software Developer — Avasoft, Chennai
Mar 2025 – Jan 2026
Client: Zywave
- Owned the ICP (Ideal Customer Profile) data-management module end-to-end on a FastAPI + PostgreSQL stack — schema design, validation, and cross-team API contracts — supporting both manually created and recommendation-engine-generated profiles.
- Built field-level change-detection by diffing incoming Pydantic DTOs against existing SQLAlchemy ORM models, enabling precise partial updates and reliable state synchronization.
- Integrated AWS Boto3 to auto-generate descriptive, AI-powered ICP names from data attributes at creation time, before persistence to PostgreSQL.
- Implemented soft-delete and user-scoped retrieval APIs returning only active, non-deleted ICPs, preserving data recoverability and enforcing per-user access boundaries.
- Partnered with frontend and technical leads on ER diagrams, schemas, and data contracts — orchestrating cross-team Search API calls under strict conditional rules, tracked via Jira.
Client: Liberty Home Guard Solution Design
- Authored the end-to-end AWS solution design for an AI-powered QA automation platform that scales agent-call review from a small manual sample to ~1,000 calls analyzed daily, replacing sampling-based QA with full-coverage automated scoring.
- Designed the knowledge-base creation flow — ECS orchestrated via EventBridge, Amazon Textract for document extraction, and Amazon Bedrock to structure QA scorecards into RDS PostgreSQL schemas for consistent cross-department scoring.
- Architected the real-time QA workflow (Amazon Connect → Transcribe with PII redaction → Step Functions → Bedrock), gated by a confidence check that auto-stores high-confidence results and routes the rest to manual review.
- Led a comparative model evaluation of Claude 3.7 Sonnet, Claude 4 Sonnet, and Nova Premier on accuracy and latency, recommending Claude 4 Sonnet for the strongest accuracy-to-latency balance.
- Delivered a start-to-finish reference architecture with per-service cost modeling, connecting business pain points to a build-ready, AWS-native design.
Projects
StudyBuddy — RAG Study Assistant
Mar 2026 – Apr 2026
FastAPI · LangChain · Groq AI · PostgreSQL (pgvector) · Docker · React Native
- Built a full-stack RAG-powered study chatbot (React Native + Expo frontend, FastAPI backend) that lets users upload PDFs up to 50 pages and ask context-aware questions grounded in the attached document.
- Engineered the retrieval pipeline with LangChain, Groq AI, and Google Embeddings, storing document vectors in PostgreSQL (pgvector) via Supabase for fast semantic search.
- Designed 18 REST endpoints across three modules — document lifecycle, chat, and quiz engine — with auto-generated Swagger docs.
- Implemented secure authentication using Supabase Auth with a dual-table user model and access-token propagation for per-user data scoping.
- Built a dynamic quiz engine that parses documents via a hybrid text + OCR pipeline (RapidOCR) and offloads MCQ generation to a background worker, with difficulty-tunable questions and async status polling.
- Implemented instant MCQ grading against database records using an optimized answer lookup map. Containerized with Docker.
Education
B.Tech, Artificial Intelligence & Data Science
Bannari Amman Institute of Technology, Erode
CGPA: 8.55
2021 – 2025
HSC
Kongu Vellalar Matric. Hr. Sec. School, Coimbatore
93.96%
2020 – 2021
SSLC
Kongu Vellalar Matric. Hr. Sec. School, Coimbatore
94.6%
2018 – 2019