DEVI SRI CHARAN VALUPADASU

Software Engineer • Scalable Backend Systems • Applied AI/ML • Full-Stack

vdevisricharan@gmail.com • • github.com/vdevisricharan • linkedin.com/in/vdevisricharan

Professional Summary

Software Engineer with over 2 years of experience building scalable backend architectures, distributed background pipelines, and production AI/LLM applications. Proficient in engineering cross-platform mobile and web services using React Native, Next.js, Django, and FastAPI, orchestrating multi-agent RAG pipelines with LangGraph and Qdrant, and automating large-scale enterprise data workflows with Celery, Redis, and AWS. Proven track record of delivering resilient, high-throughput systems and data quality pipelines across enterprise and startup environments.

Work Experience

Software Engineer Skizen • Hyderabad, India
  • Architected and developed a mobile-first social discovery platform connecting users through location-based demand mapping, engineering the cross-platform client with React Native (Expo) and backend services using Django, DRF, Express, and PostgreSQL with GeoJSON 2dsphere spatial indexing for business discovery, category feeds, and tiered RBAC.
  • Engineered high-performance video feed orchestration using expo-video and viewport visibility detection (onViewableItemsChanged), guaranteeing single-stream hardware-accelerated playback with preloading, zero black-frame transitions, and non-blocking background media ingestion.
  • Developed a production-quality local video automation tool (Python, OpenCV, PySceneDetect, FFmpeg) to prepare raw footage for 9:16 Instagram Reel editing; implemented automated scene cut detection, visual quality scoring (sharpness, motion), perceptual duplicate filtering, and programmatic CapCut Desktop draft project generation with non-destructive speed retiming (1.0x, 0.75x, 0.5x slo-mo) and an interactive human-in-the-loop review UI.
  • Engineered an automated social media posting pipeline that monitors client Google Drive folders for finalized Reels and automatically schedules and publishes content across client social media handles (Meta Graph API / Instagram Reels, YouTube Shorts) with chunked media upload streaming, OAuth token lifecycle management, rate limiting, and resilient retry mechanisms.
  • Built asynchronous backend pipelines using Celery and Redis to decouple background analytics aggregation and push notifications; implemented bidirectional real-time messaging using Socket.IO with JWT authentication, typing indicators, and TOTP-based two-factor authentication (2FA).
Software Engineer Lumiq • Noida, India
  • Engineered an AI-powered virtual sales agent utilizing LangChain and LangGraph to orchestrate stateful, multi-step LLM conversational workflows, coupled with a Next.js (SSR/SSG) frontend enabling real-time interactive product demonstrations for enterprise prospects.
  • Designed and implemented a high-throughput data deduplication engine using Python, Django, and AWS ETL pipelines (S3, Glue, Athena) to identify, merge, and clean duplicate insurance policy records across enterprise data lakes.
  • Developed and executed an automated Bash and SQL data correction script that successfully eliminated over 350,000 duplicate financial exposure records for a key insurance client, ensuring regulatory data reliability and financial reporting accuracy.
  • Built a scalable internal project management and resource allocation platform using React, NestJS, and PostgreSQL, streamlining cross-team project tracking and optimizing operational resource utilization.

Technical Projects

Multi-Agent Medical Assistant [github.com/vdevisricharan/Multi-Agent-Medical-Assistant]
Python FastAPI LangGraph Qdrant Docling PyTorch Docker
  • Architected a modular multi-agent medical assistance system using LangGraph and FastAPI, dynamically routing complex healthcare research and triage queries across specialized conversational, literature retrieval, and medical imaging agents.
  • Engineered an advanced agentic RAG pipeline integrating Docling for multimodal document parsing (extracting tables, text, and figures), LLM query expansion, and Qdrant hybrid search (BM25 sparse keyword + dense embeddings) with HuggingFace Cross-Encoder (ms-marco-TinyBERT) re-ranking.
  • Developed deep learning computer vision agents in PyTorch for chest X-ray disease classification and skin lesion segmentation, incorporating input/output safety guardrails, confidence-based agent handoffs (RAG to web search), and human-in-the-loop verification workflows.
  • Containerized the multi-service architecture with Docker and automated continuous integration testing with GitHub Actions; integrated ElevenLabs API for low-latency bidirectional voice interaction.
Cost-Efficient RAG Application [github.com/vdevisricharan/cost-efficient-rag-application]
Python FastAPI Gemini 2.5 Flash LanceDB ChromaDB SentenceTransformers
  • Engineered a self-hosted, serverless-style RAG service pairing embedded vector stores (LanceDB with Apache Arrow columnar disk format and ChromaDB) with Gemini 2.5 Flash, demonstrating up to 99.9% analytical infrastructure cost reduction compared to always-on managed cloud vector pods.
  • Built an idempotent multi-format document ingestion pipeline (PDF, HTML, Markdown) utilizing recursive chunking, SHA-256 chunk hashing to eliminate duplicate vector writes, and strict system-prompted citation enforcement [Doc: <source>, Chunk: <id>].
  • Implemented a configurable cosine similarity cutoff (τ = 0.35) achieving 100% fallback accuracy on unanswerable queries in benchmark tests; evaluated retrieval performance across a 20-query ground-truth dataset yielding 83.3% Recall@3, 83.3% Hit Rate@3, and 85.0% LLM-as-a-Judge answer relevance.
  • Exposed FastAPI REST endpoints (/ingest, /query, /health, /stats) and CLI tooling; instrumented structured logging via Loguru to monitor per-query retrieval latency (p50 = 26.48 ms on LanceDB), token consumption, and generation metrics.

Technical Skills

Programming: Python, JavaScript, TypeScript, SQL, Java, Bash
Frontend & Mobile: React, React Native, Next.js, Redux, Zustand, Tailwind CSS, HTML5, CSS3
Backend & APIs: Django, Django REST Framework, FastAPI, Flask, Node.js, Express.js, NestJS, REST APIs, Socket.IO
AI / ML / RAG: LangChain, LangGraph, PyTorch, TensorFlow, Hugging Face, Transformers, Qdrant, LanceDB, ChromaDB, Docling, OpenCV, SentenceTransformers, RAG, NLP
Data, Cloud & DevOps: PostgreSQL, MySQL, MongoDB, AWS (EC2, S3, Lambda, Athena, Glue), Celery, Redis, Docker, Git, GitHub Actions, Firebase / FCM, Cloudinary

Education

Indraprastha Institute of Information Technology Delhi (IIIT Delhi)
B.Tech in Computer Science and Design • 2020 – 2024
Coursework: Data Structures & Algorithms, DBMS, Advanced Programming, Analysis & Design of Algorithms, Computer Networks, Operating Systems, NLP, Information Retrieval, Machine Learning, Reinforcement Learning in Finance, Affective Computing.

Achievements & Certifications

Honors & Awards
• Won 1st Place | SMAI Hackathon by Micron Technology India (2022)
• Scored 99.23 Percentile | JEE MAINS (All India Rank: 8,694)
Certifications
• AWS Certified Data Engineer – Associate (Partner) | Lumiq
• AWS Certification | Scaler