Hi, I'm Shubham
I'm a Software Engineer based in Washington, DC, passionate about creating intelligent, user-centered systems using Machine Learning and modern software tools.
Currently seeking full time opportunities
About Me
Hi, my name is Shubham, and I'm a passionate Software Engineer and problem solver currently pursuing my studies at George Washington University. I have a strong foundation in Full-Stack Development, Android Development, and AI Development. I'm actively seeking opportunities where I can contribute to building secure, scalable, and user-focused tech solutions.
I specialize in crafting web and mobile applications using technologies like JavaScript, TypeScript, React, Node.js, Kotlin, Python, Firebase, and AWS cloud services. I enjoy developing intuitive user experiences, writing clean code, and building infrastructure that scales.
I thrive in collaborative and fast-paced environments, whether I'm developing secure APIs, enhancing backend performance, building Android applications, or exploring intelligent systems through AI development. I'm also diving deeper into Machine Learning to create data-driven, impactful solutions that can learn, adapt, and solve real-world problems.
Let's connect and build something amazing together!
Experience
Software Engineer @ Zydus
Aug 2019 - Nov 2023
Java, Spring Boot, Python, Node.js, React, TypeScript, Next.js, PostgreSQL, DynamoDB, AWS Lambda, Step Functions, SQS, X-Ray, CloudWatch, Jest, PyTest
Overview
- Owned the end-to-end delivery of scalable software solutions, bridging the gap between high-performance backend microservices, intelligent data processing, and complex user interfaces.
Backend Architecture & Distributed Systems
- ▹Eliminated downtime during 3x traffic spikes (billions of requests) by decoupling a legacy monolith into event-driven microservices using Java (Spring Boot) and AWS Lambda, ensuring system stability at scale.
- ▹Reduced P99 data retrieval latency by 50% (from 200ms to <100ms) by optimizing PostgreSQL schema designs and DynamoDB indexing strategies for critical high-volume services.
- ▹Reduced authentication incidents by 25% by designing scalable REST and GraphQL APIs using Node.js and Python, implementing robust rate-limiting and circuit-breaker patterns.
- ▹Ensured zero-downtime reliability for billing systems by building event-driven orchestration layers using AWS Step Functions and SQS to process usage metering in real-time.
AI Infrastructure & Data Engineering
- ▹Enabled downstream predictive analytics and anomaly detection by engineering high-throughput data pipelines using Python (Pandas/NumPy) to transform massive telemetry datasets into ML-ready streams.
Frontend Engineering & UX
- ▹Improved reporting workflow efficiency by 30% for 5,000+ daily enterprise users by engineering a full-stack analytics platform with complex data grids using React, TypeScript, and Next.js.
- ▹Minimized Time to First Byte (TTFB) and achieved WCAG 2.1 AA compliance by implementing Edge Caching and Server-Side Rendering (SSR) in Next.js, expanding accessibility for diverse user bases.
Leadership & DevOps
- ▹Decreased Mean Time to Resolution (MTTR) for production incidents by 50% by implementing system-wide observability and distributed tracing using AWS X-Ray and custom CloudWatch metrics.
- ▹Reduced production rollout failures by 35% by architecting internal CI/CD workflows and automated testing suites that achieved 95% code coverage using Jest and PyTest.
- ▹Reduced deployment rollbacks by 30% by serving as a Technical Mentor for the engineering team, conducting rigorous full-stack code reviews and standardizing architectural documentation.
Projects
I'm building a project called VoiceSense, a real-time emotion detection system that analyzes voice input to classify emotional states like happiness, sadness, or anger using deep learning models. I used Python, PyTorch for model training, Librosa for audio feature extraction, and Coqui TTS to generate voice responses that match the detected emotion. The system also uses FastAPI for backend inference and a React-based frontend where the tile color changes dynamically based on the emotion detected. I'm currently developing a new feature to bring this into live video calls, where the system will process voice in real time and update the participant's tile color during the call to reflect their emotional state.
A full-stack AI chatbot platform built from the ground up with conversational memory and intelligent code analysis. Features include follow-up question support using LangChain, automatic routing between chat and code improvement pipelines, and real-time subscription management with Stripe integration. Built with Next.js, TypeScript, and Tailwind CSS frontend, Node.js/Express backend, PostgreSQL with Prisma ORM, and Clerk authentication. The AI engine runs locally using Ollama with LLaMA 3, eliminating dependency on external APIs. Includes tiered subscription plans (Basic, Pro, Plus) with usage limits and instant UI updates. Demonstrates advanced AI integration with production-ready features like JWT security, real-time data tracking, and seamless user experience.
Production-grade quality & safety layer for LLM/RAG answers, built to Google Search/Gemini standards. Intercepts outputs and routes ALLOW / REPAIR / BLOCK via faithfulness, coverage, and safety (PII/toxicity) scores; auto-repairs with grounded citations. Features hybrid retrieval (BM25 + TF-IDF, RRF), policy-driven YAML configuration, shadow mode analytics, and optional semantic faithfulness. Fully observable with Prometheus (p95 < 500ms) and Jaeger tracing, Dockerized, and ships with a TypeScript SDK, golden tests, and a 50-case benchmark. Built with Python/FastAPI, Redis/RQ, NumPy/scikit-learn. Results: reduced hallucinations by 30–50%, increased citation faithfulness by 20–35% a guardrail tailored for Google-level answer quality.
What's Next
Get In Touch
I'm currently looking for Full Time jobs.
Whether you have a project to discuss or just want to say hi, my inbox is open for all!
Designed and built by Shubham Avhad