AI Engineer with 1+ years in transformer models, RAG & MLOps
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
AI Engineer with hands-on experience fine-tuning transformer models (ViT, DeiT) and building production RAG systems. Deployed Dockerized FastAPI inference services on AWS EC2 at 120 ms median latency and implemented automated ETL, CI/CD, and MLflow experiment-tracking workflows. Comfortable writing PyTorch training loops from scratch, working with the Hugging Face ecosystem (transformers, PEFT, datasets), and quantising models to INT8/6-bit/4-bit for edge and cost-efficient deployment. Looking to apply these LLM and MLOps skills to fine-tune open-weight models on educational datasets and build evaluation pipelines that close the gap between a model and something genuinely useful.
Indian Institute of Technology, Jodhpur
B.Tech · Electrical Engineering
August 1, 2022 – June 30, 2026
Sri Chaitanya Junior College
Senior Secondary · XII
June 1, 2021 – May 31, 2021
Sri Chaitanya Techno
Secondary · X
June 1, 2019 – May 31, 2019
Bluestock
AI/ML Intern
May 1, 2025 – July 31, 2025
India
Vision Transformers
July 1, 2025 – October 31, 2025
Implemented ViT and DeiT architectures from scratch in PyTorch on a drone-imagery dataset, writing custom training loops and data pipelines for high-throughput image analysis. Conducted multi-level quantisation (Float32 → INT8 / 6-bit / 4-bit) using custom NumPy/PyTorch logic, extracting and validating 60+ weight tensors across 12 encoder layers - achieving a 3× reduction in model size with <2% accuracy loss to enable cost-efficient edge deployment. Benchmarked low-precision inference across precision levels to compare throughput, memory footprint, and accuracy trade-offs — directly analogous to AWQ/GPTQ quantisation workflows.
PolicyPilot
January 1, 2025 – June 30, 2026
Built a transformer-based RAG pipeline on internal policy documents using Hugging Face and FAISS; implemented document chunking, vector embeddings, and semantic retrieval to ground responses and measurably reduce hallucinations over a keyword-search baseline. Integrated confidence-based human-in-the-loop escalation for low-confidence responses and deployed the service as a Dockerized FastAPI endpoint with structured logs and full response traceability.
DocuAssist
January 1, 2025 – June 30, 2026
Developed a private RAG chatbot over PDFs and internal documents with citation-backed responses and confidence scoring to reduce hallucinations. Designed and evaluated multiple chunking and embedding strategies to improve retrieval relevance; added source citations per answer to improve reliability and user trust.
Music Genre Classification
January 1, 2024 – May 31, 2024
Evaluated multiple classification algorithms on 10k+ audio files; achieved 90% accuracy using a Decision Tree + AdaBoost ensemble after systematic hyperparameter search.
Cultural Fit Analysis
The candidate's academic projects, personal projects, and internship experience demonstrate a strong interest and practical application in AI/ML, aligning well with an AI Engineer role. The diversity of projects, from vision transformers to RAG chatbots and ETL pipelines, shows a broad technical curiosity and willingness to tackle different challenges. The involvement in various positions of responsibility at IIT Jodhpur suggests leadership potential and engagement in extracurricular activities, contributing positively to cultural fit.
Soft Skills & Operational Fit
The candidate's project descriptions indicate a structured approach to problem-solving, including systematic hyperparameter search, evaluation of multiple strategies, and benchmarking. The integration of human-in-the-loop escalation and structured logging suggests an understanding of operational reliability and user trust. Participation in hackathons and mentorship roles indicates initiative and collaboration potential.