Hi, I'm
Prashant Bastola
Building end-to-end language AI systems, instruction tuning, RAG pipelines, and low-latency production AI inference services.
About Me
Prashant Bastola
LLM Engineer & NLP Researcher
Professional Summary
Results-driven LLM Engineer and NLP Researcher specializing in end-to-end language AI systems, particularly for low-resource languages like Nepali.
Proven expertise in Generative AI, Large Language Models (LLMs), instruction tuning, Retrieval-Augmented Generation (RAG), tokenizer adaptation, and transformer-based NLP.
Strong academic and research background with IEEE publications in OCR and legal NLP, paired with hands-on production experience building scalable, low-latency AI inference services using PyTorch, FastAPI, Docker, and AWS.
2 IEEE Publications
OCR & Agricultural AI research papers
Zeta Labs ML Intern
Sub-100ms AWS inference & RAG resume parser
IIT Kanpur SURGE Intern
Legal coreference models & 10k+ legal corpus chunking
8.41 / 10.0 CGPA
MCA in AI & ML (Chandigarh University)
8.68 / 10.0 CGPA
BCA in Cloud Computing (Gujarat University)
NepaliLLM Research
50k+ instruction tuning & 4-bit QLoRA fine-tuning
Work Experience
Machine Learning Intern
Zeta Labs
- Deployed robust inference pipelines on AWS using FastAPI and Docker, successfully serving Transformer models with sub-100ms latency.
- Architected and deployed a production-grade RAG resume screening system utilizing Pinecone vector embeddings and custom chunking strategies, processing over 5,000+ applications and reducing manual candidate review time by 20% (saving ~15 hours weekly).
- Integrated NLP intent classification for system commands by leveraging LLM function calling, structured output validation, and implementing strict fallback guardrails.
Research Intern
IIT Kanpur (SURGE)
- Built and fine-tuned coreference models for legal text using PyTorch and Hugging Face, achieving a 15% improvement in entity-linking F1-scores.
- Scripted custom spaCy training loops and parsing algorithms to ingest and chunk a 10,000+ document unstructured legal corpus with 90% accuracy.
- Resolved complex, long-range entity-linking challenges within large legal corpora, significantly optimizing retrieval accuracy and semantic context alignment.
Education & Degrees
Masters of Computer Applications (MCA)
Major in Artificial Intelligence & Machine Learning
Chandigarh University
Punjab, India
• Advanced coursework in Deep Learning, Natural Language Processing, Reinforcement Learning, and Markov Decision Processes.
• Developed Agentic RAG Chatbot Systems using LangChain and pgvector across 500+ documents.
Bachelor of Computer Applications (BCA)
Minor in Cloud Computing
Gujarat University
Ahmedabad, Gujarat, India
• Specialized in Cloud Infrastructure, Distributed Data Processing, and Software Engineering principles.
• Graduated with High Academic Distinction (8.68 / 10.0 CGPA).
Featured Projects
Directly from resume: Instruction-tuned LLMs, Agentic RAG Chatbots, and Computer Vision OCR pipelines.
Face Recognition Attendance System
2026
Developed a real-time face recognition system using Python and OpenCV for attendance and access control.
Technologies
Agentic RAG Chatbot
2025
Built a Retrieval-Augmented Generation (RAG) chatbot using LangChain and FAISS for intelligent document querying.
Technologies
TwinLight-MARL
2026
Developed a multi-agent reinforcement learning framework for intelligent traffic signal control using PyTorch and GAT.
Technologies
Research Publications
Directly from resume: Peer-reviewed research papers in OCR and LLM agricultural advisory systems.
Unified Framework for Context-Aware OCR
Prashant Bastola, et al.
Summary: A novel hybrid deep learning architecture combining Spatial Transformer Networks (STN) with Transformer-based language modeling to perform high-precision character recognition on low-resource script documents.
A Machine Learning and LLM-Based System for Crop Recommendation, Revalidation, and Fertilizer Advisory
Prashant Bastola, et al.
Summary: Investigates the integration of ensemble decision trees and fine-tuned Large Language Models for automated soil composition analysis, crop yield advisory, and real-time agricultural revalidation.
Skills & Proficiencies
Directly from resume: LLM Engineering, NLP frameworks, Retrieval-Augmented Generation, and MLOps.
Programming Languages
AI & Machine Learning
LLM Engineering
NLP & Devanagari Processing
Retrieval & Search (RAG)
Deployment & MLOps
Data & Visualization
GitHub Activity
Consistent open-source contributions, repository stars, and active maintenance of AI toolkits.
Pinned Repositories
Instruction Tuned LLM for Low-Resource Nepali Language (7B Params, QLoRA)
Spatial Transformer Networks coupled with Transformers for degraded script OCR
Cultural Archival & Multimodal AI dialect translation platform
NLP vector matching & ATS score matrix for candidate profiling
Contact Prashant Bastola
Open for LLM Engineer, NLP Researcher, and Machine Learning engineering roles. Reach out directly!
Contact Details
Professional Profiles
Send a Direct Message
Fill out the form below to discuss opportunities or inquiries.