Skip to content
View asthasingh0660's full-sized avatar
🦄
🦄

Block or report asthasingh0660

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
asthasingh0660/README.md

Astha Singh

Computer Engineering Undergraduate  ·  AI/ML Engineer  ·  Full-Stack Developer

Building production-grade AI systems, cloud-native pipelines & distributed backends — end to end.

LinkedIn GitHub Portfolio Email


About Me

  • B.Tech in Computer Engineering @ Bharati Vidyapeeth (Deemed University), Pune — CGPA: 9.08 (May 2026)
  • Experienced in end-to-end LLM applications — RAG pipelines, agent-to-agent orchestration, MCP-style tool routing
  • Built cloud-native systems on GCP (Cloud Run, BigQuery, Cloud Scheduler) within free-tier constraints
  • Sole Frontend Engineer across 3 production SaaS products at Floww APIs Pvt Ltd
  • Project Intern @ Indian Air Force (9BRD) — AI-based PCB diagnostic system
  • Deployed production apps with ~95% accuracy, ~2-3s latency, at $0/month infrastructure cost
  • Open to full-time opportunities in India and abroad

Tech Stack

AI & LLMs

Python LangChain RAG Pipelines Prompt Engineering Scikit-learn XGBoost

Backend & Systems

FastAPI Node.js Flask Redis PostgreSQL Docker Nginx

Cloud & Infra

GCP Vercel Render Linux Git

Frontend & Tools

React Streamlit Figma Postman SQL


Featured Projects

FastAPI React Node.js Bull Redis PostgreSQL Docker Nginx 5 cloud platforms

LLM-powered code transpilation across 11+ languages using a distributed async architecture with Bull + Redis job queue separating API ingestion from LLM inference — enabling non-blocking requests and horizontal worker scaling.

  • ~95% syntax correctness · ~2-3s latency · $0/month infrastructure cost · 99.5%+ uptime
  • SHA-256 content-addressed caching with 45% cache hit rate, reducing inference cost to ~$0.003/request
  • Judge0 sandbox execution for automated compile-time output validation
  • Deployed across 5 independent platforms with full system design documentation

Python LangChain Whisper ElevenLabs Streamlit

Voice-first RAG support agent over HuggingFace documentation using Whisper (STT) and ElevenLabs (TTS), with hybrid semantic + keyword retrieval and multi-LLM routing.

  • Retrieval accuracy improved from ~50% to ~70% (~40% relative improvement) on real troubleshooting queries
  • Multi-LLM routing (Groq, Gemini, OpenAI) with session-based conversational memory
  • ~50% reduction in follow-up clarifications · ~60% shorter spoken output for voice UX

Python GCP Cloud Run BigQuery Cloud Scheduler Docker Flask Claude API

Cloud-native pipeline ingesting real-time weather sensor data into BigQuery via containerized Cloud Run, with statistical anomaly detection and AI-generated human-readable insights.

  • Z-score anomaly detection integrated with Claude API for interpretable alert summaries
  • Entire system designed within GCP free-tier limits, demonstrating cost-efficient infrastructure from scratch
  • Live monitoring dashboard visualizing anomalies, sensor activity, and real-time metrics

Python Scikit-learn XGBoost SHAP Streamlit

Fairness-aware ML pipeline surfacing the accuracy-fairness tradeoff in HR attrition prediction, built for non-technical decision makers.

  • Compared 4 models (LR, Random Forest, XGBoost, MLP) on a 5.2:1 imbalanced HR dataset
  • Quantified AUC-bias correlation = 0.90 · Applied Reweighing mitigation to reduce gender disparity
  • 6-tab Streamlit dashboard with SHAP explainability + fairness auditing across 3 metrics

Experience

Frontend Engineering Intern — Floww APIs Pvt Ltd, Pune Jun 2025 – Jan 2026

Sole frontend engineer across 3 production SaaS products in the Atom Suite ecosystem (Atom HR, Atom Mail, Nucleus). Built full React architecture with RBAC across Employee, Manager, and Admin roles, Firebase OTP auth, GPS attendance, task & project tracking, and billing management — from scratch through production deployment.

Project Intern — Indian Air Force (9BRD), Pune Jun 2025 – Jul 2025

Led Windows-to-Linux migration for an AI-based PCB diagnostic system. Configured a secure Linux stack supporting AI workflows and collaborated with defence R&D engineers to validate setup and support production-ready prototype deployment.

UI/UX Developer Intern — Intertwine Networks Pvt Ltd Jul 2024 – Nov 2024

Designed authentication flows, interactive dashboards, and referral analytics tools. Delivered high-fidelity Figma prototypes and contributed to front-end implementation across Agile sprints.


Achievements

  • Next Generation Tech Booster Program Scholarship
  • Advanced to Prototype Round — Amdocs GenAI Hackathon
  • Qualified Smart India Hackathon (SIH) — College-level rounds (twice)
  • 10+ certifications from Google, IBM, Meta, Coursera & NPTEL across AI, ML, Cybersecurity, and Databases
  • Head of Design & Branding @ ACES & Financity — boosted event participation by 20%

GitHub Stats


"Define the problem. Measure what matters. Ship. Iterate."

Open to full-time roles in AI Engineering, Backend, or Full-Stack Development — India & abroad.

asthasingh0660@gmail.com

Pinned Loading

  1. DevTranspiler DevTranspiler Public

    DevTranspiler is a production-grade AI code transpilation platform with async job queue, intelligent SHA-256 caching (45% hit rate), and free-tier multi-cloud deployment. FastAPI + React + Bull + R…

    Python

  2. HuggingFace_voice_customer_support_agent HuggingFace_voice_customer_support_agent Public

    Built a voice-enabled conversational RAG support agent with hybrid retrieval, multi-LLM orchestration, Whisper STT, and ElevenLabs TTS, improving grounded response accuracy (~40%) and reducing clar…

    Python

  3. SmartPark SmartPark Public

    AI-powered Smart Parking & Route Planning system using A* and Dijkstra with real-time traffic prediction and dynamic spot allocation.

    Java

  4. responsible-ai-attrition-dashboard responsible-ai-attrition-dashboard Public

    Responsible AI dashboard for employee attrition analysis using ML, fairness metrics, and explainability

    Python

  5. Farmer_Management_System Farmer_Management_System Public

    Full-stack Farm Management System built with Flask and MySQL. Includes farmer registration, agro-product management, secure login, and SQL triggers for automated logging. Demonstrates strong SQL, b…

    CSS

  6. sentinel-anomaly-detector sentinel-anomaly-detector Public

    Real-time AI anomaly detection pipeline - streams live weather data from 5 global cities into BigQuery, runs z-score detection, and uses Claude AI to explain anomalies in plain English. Deployed on…

    HTML