--:--:--

SHIVAM GOEL

Shivam Goel
shivam@dtu
OS: Ubuntu 24.04.3 LTS
HOST: New Delhi, India, Earth
DEGREE: B.Tech, Mechanical Engineering (2025 – 2029)
ORG: Delhi Technological University (DTU)
ROLE: AI/ML Engineer and SDET Developer
EXPERIENCE: Autoload AI Systems (SDET) | Veracy (AI/ML)
FLAGSHIPS: HearUsOut (19+ dB SI-SDRi), Andrew Ng RAG, Agentic AI
SKILLS: Python, PyTorch, LangChain, FAISS, Groq, C, JS
CERTS: 6x DeepLearning.AI / Coursera Specializations
EMAIL: shivamgoyal1217@gmail.com
GITHUB: github.com/sh1vam-03
LINKEDIN: linkedin.com/in/shivam-goel-73a188362
STATUS: Open to collaborate & opportunities
Try: about · projects · experience · skills · certs · socials · resume · help
shivam@dev:~$
New Delhi, India • Open to collaborate

Shivam
Goel

AI/ML Engineer & Researcher

Building scalable, intelligent systems. Specializing in Deep Learning Speech Separation, Agentic Workflows, and High-Reliability SDET Automation.

View résumé GitHub LinkedIn Email
Shivam Goel
~/shivam
SCROLL
WORK

Where I've worked

SDET Intern (Software Development Engineer in Test)

2026
Autoload AI Systems Private Limited
  • Gained hands-on experience in software development, testing, automation, and performance scripting with a focus on reliable and scalable software.
  • Worked with development and testing workflows to improve software quality, reliability, and performance across production services.

AI/ML Intern

2026
Veracy Consulting
  • Worked on Python programming, comprehensive data preprocessing, exploratory data analysis (EDA), and supervised and unsupervised learning.
  • Developed and evaluated machine learning models for practical AI/ML applications and business datasets.
PROJECTS

Things I've built

HearUsOut

Real-World Speech Separation • 19+ dB SI-SDRi

Speech separation pipeline using fine-tuned SR-CorrNet, achieving 19+ dB SI-SDRi on noisy audio. Designed adaptive preprocessing and VAD routing for clipping, silence, and overlapping speakers with zero-shot extraction.

Python PyTorch SR-CorrNet DeepFilterNet ECAPA-TDNN

Andrew Ng Digital Twin

Conversational RAG AI with Voice I/O

RAG-powered conversational AI grounded on Andrew Ng lecture notes. Implemented Gemini/Groq fallback, short/long-term memory, automated citations, and offline voice I/O in Streamlit.

Python LangChain FAISS Gemini Groq Streamlit

Agentic AI Research System

Autonomous Research Planning & Verification

Multi-agent research system for autonomous planning, retrieval, generation, and verification of arXiv research. FAISS vector retrieval with bge-small-en-v1.5 embeddings and iterative self-reflection.

Python Agentic AI FAISS Groq Ollama RAG

Stock Market Web Application

LSTM Price Prediction & News Chatbot

Collaborated in a team of 4 to build a full-stack stock market prediction application. Implemented LSTM-based price forecasting, real-time news aggregation, and an interactive AI chatbot.

Python LSTM Machine Learning Web Dev
SKILLS

Technical Stack

LANGUAGES
Python C JavaScript Bash
ML & AI
PyTorch TensorFlow Scikit-learn Computer Vision OpenCV NLP XGBoost pandas NumPy
AI & LLMs
LangChain FAISS RAG Agentic AI Multi-Agent Systems Groq Gemini API Ollama
BACKEND & SDET
Test Automation Performance Scripting CI/CD Workflows Streamlit
HARDWARE & TOOLS
ESP32 Fusion 360 Arduino VS Code Git
INSTITUTION
Delhi Technological University B.Tech Mechanical '29 New Delhi
CERTIFICATIONS

Verified Credentials

Deep Learning Specialization
Coursera • DeepLearning.AI
Machine Learning
Coursera • DeepLearning.AI
Improving Deep Neural Networks: Hyperparameter Tuning
Coursera • DeepLearning.AI
Structuring Machine Learning Projects
Coursera • DeepLearning.AI
Sequence Models
Coursera • DeepLearning.AI
Machine Learning in Production
Coursera • DeepLearning.AI
DOMAINS & IMPACT

THE ENGINEERING
SYSTEM

A comprehensive overview of applied machine learning pipelines, autonomous agents, and reliability engineering.

01

Speech Separation

Fine-tuned SR-CorrNet achieving 19+ dB SI-SDRi with ECAPA-TDNN zero-shot target extraction.

02

Agentic AI Systems

Autonomous arXiv research planning, FAISS vector retrieval, and Groq/Ollama tool-calling.

03

SDET Automation

Hands-on software development, automated test regression suites, and performance scripting.

04

Conversational RAG

Andrew Ng Digital Twin with dual LLM fallback, persistent memory, and voice I/O in Streamlit.

05

Predictive ML Models

LSTM-based stock forecasting, exploratory data analysis, and supervised machine learning.

06

Embedded Hardware

ESP32, Arduino microcontrollers, circuit prototyping, and 3D modeling with Fusion 360.

01

Speech Separation

Fine-tuned SR-CorrNet achieving 19+ dB SI-SDRi with ECAPA-TDNN zero-shot target extraction.

02

Agentic AI Systems

Autonomous arXiv research planning, FAISS vector retrieval, and Groq/Ollama tool-calling.

03

SDET Automation

Hands-on software development, automated test regression suites, and performance scripting.

04

Conversational RAG

Andrew Ng Digital Twin with dual LLM fallback, persistent memory, and voice I/O in Streamlit.

05

Predictive ML Models

LSTM-based stock forecasting, exploratory data analysis, and supervised machine learning.

06

Embedded Hardware

ESP32, Arduino microcontrollers, circuit prototyping, and 3D modeling with Fusion 360.