Prateek Goswami
Building Intelligent Systems,
LLM Applications & AI Agents
B.Tech CSE (AI) • 9.09 CGPA • IIT Gandhinagar & IIT Jammu Exchange Scholar
Building AI Systems That
Solve Real Problems
Computer Science & AI undergraduate specialising in Machine Learning, Computer Vision, LLM Applications, and Agentic AI Systems. I build production-oriented AI solutions — from intelligent chatbots and multimodal models to large-scale ML pipelines and autonomous AI agents.
I am a Computer Science & AI undergraduate with hands-on experience building end-to-end ML pipelines, AI-powered tools, and multimodal systems across NLP, Computer Vision, and large-scale data engineering.
My expertise spans Python, agentic AI workflows, prompt engineering, and deep learning across real-world datasets and production-grade tasks.
Seeking opportunities in AI Engineering, Machine Learning, Agentic AI, and Applied Data Science.
My focus is not just training models, but deploying AI systems that interact with users, tools, and real-world workflows.
9.09 / 10
B.Tech CGPA
Top Academic Performance
7+
AI Projects
End-to-end ML & CV Systems
2
IIT Exchange Programs
IIT Gandhinagar & IIT Jammu
Python / AI Agents
Key Proficiencies
Applied Tool Use & LLMs
Professional Experience
AI / ML Engineer Intern
Patidar Agriculture
Built and deployed an AI-powered chatbot for the company website to handle product queries, machinery recommendations, and customer support — integrating an LLM backend with a conversational interface.
Developed computer vision models for disease detection using image processing and predictive models for data-driven decision support using Python, OpenCV, and Scikit-learn.
Worked across the full ML lifecycle — data collection, preprocessing, model training, evaluation, and deployment — delivering end-to-end solutions for a production website used by real customers.
Technical Proficiencies
Languages & Web
AI & Agentic Systems
ML / Deep Learning
Computer Vision & NLP
Data Science
Tools & Ecosystem
Featured Projects
AI Outreach Agent
Built and deployed a LangGraph-based AI agent with tool-calling workflows that scrape company websites, score resume-company fit, generate personalized outreach emails, and evaluate outputs through hallucination detection and quality-scoring pipelines.
- Tested and debugged agent outputs to improve grounding and reduce hallucination-prone responses.
- Designed a human-in-the-loop review workflow with retrieval-based Style Memory, enabling future generations to adapt from previous user edits and approved responses.
- Integrated SQLite, Google Drive API, and Google Sheets API for persistent tracking of outreach history, evaluation metrics, approval decisions, and generated communications across the agent lifecycle.
Real-Time Anti-Cheat Proctoring
A real-time multi-model inference system detecting gaze deviation, head pose, hand gestures, and unauthorized devices.
- Logged behavior events with timestamped alerts for fully automated exam monitoring constraints.
- Orchestrated multiple CV models in a highly efficient single pipeline showcasing applied agentic design.
Visual Question Answering (VQA)
A multimodal neural network generating natural language answers based on images and corresponding questions.
- Combined ResNet/ViT feature extraction with BERT question encoding structures.
- Evaluated on VQA v2 benchmark covering open-ended and binary question types.
Customer Churn Prediction Pipeline
An end-to-end binary classification pipeline engineered to process large-scale customer records.
- Analyzed 7,000+ customer records and engineered features via StringIndexer and VectorAssembler.
- Trained Logistic Regression/Random Forest models; tracked experiments with MLflow, achieving 0.83 AUC-ROC.
Sentiment Analysis Pipeline
An end-to-end NLP pipeline classifying sentiments across massive textual datasets.
- Compared TF-IDF + classical ML baselines with fine-tuned transformer models.
- Reported metrics through F1-score and confusion matrix visualizations for clear business insights.
Attention Image Captioning
Developed an attention-based encoder-decoder model to automatically generate descriptive text captions for images.
- Trained on Flickr30k dataset using a custom CNN-LSTM pipeline.
- Significantly improved BLEU-4 scores through systematic attention integration.
Lip Sync Generation Pipeline
Fine-tuned Wav2Lip-based deep neural networks to synthesize realistic lip movements in video matching arbitrary audio.
- Engineered preprocessing steps covering facial land-marking, mouth region-of-interest extraction.
- Synchronized high-fidelity audio alignment streams with video frames dynamically.
Education & Academic Paths
B.Tech in Computer Science & Artificial Intelligence
JK Lakshmipat University
Selected for Academic Exchange Program
IIT Gandhinagar
Selected for Academic Exchange Program
IIT Jammu
Key Achievements
Organising Head, HackJKLU 2026
- Led operations for a national-level hackathon hosting 400+ participants, coordinating 200+ volunteers, sponsors, mentors, judges, logistics, and venue management.
- Oversaw sponsor outreach, participant onboarding, and real-time problem resolution across all hackathon tracks to deliver a seamless event experience.
Teaching Assistant - C Programming
- Selected based on academic performance and programming expertise to support 60+ students.
- Conducted structured doubt sessions, graded code assignments, and taught debugging/core C topics.
Dean’s Honour List Recipient
- Recognised for outstanding academic performance across consecutive semesters.
- Maintained top-tier marks with a consistent focus on computational fundamentals and applied AI.
Let's Build
Something Great
Have an opening for an AI Engineer role, a collaborative pipeline proposal, or just want to say hi? Connect with me directly!