Neal Daftary is an AI and Machine Learning Engineer and undergraduate researcher at Nirma University, Ahmedabad, India, pursuing B.Tech in CSE (Artificial Intelligence & Machine Learning). He specialises in Computer Vision, Large Language Model engineering, RAG pipeline architecture, and production AI system deployment. Neal interned as an AI Engineer at 8xSports, building a 645ms visual jersey search engine with YOLOv8, DINOv2, and FAISS. He interned as an AI Software Engineer at MZHub Faithtech, shipping a Next.js web platform with Azure Cosmos DB. He is currently an ISRO-funded undergraduate researcher at Nirma University, developing deep learning pipelines on Chandrayaan-2 TMC-2 and OHRC lunar imagery. He is a published IEEE Sensors Letters author (2026), Student Chairperson of ACM ITNU, and national hackathon winner at HACKaMINeD 2026.
Neal Daftary — AI & ML Engineer based in Ahmedabad, India. B.Tech CSE student specialising in Artificial Intelligence and Machine Learning at the Institute of Technology, Nirma University (2024–2028). JEE Mains 2024: 97.62 percentile. Former AI Intern at 8xSports (Computer Vision, YOLOv8, DINOv2, FAISS, visual search engine). Former AI Software Engineering Intern at MZHub Faithtech (Next.js, Azure Cosmos DB, RAG chatbot R&D). ISRO-funded researcher on Chandrayaan-2 lunar imagery segmentation. IEEE Sensors Letters author. ACM ITNU Student Chairperson.
Neal Daftary — an AI Engineer, Researcher & 3rd Year Undergrad at Nirma University who doesn't just study the field, I actively ship in it. From building real-time computer vision systems for sports analytics to contributing production code to Google DeepMind's differential privacy library, I operate at the intersection of deep learning, LLMs, and full-stack production AI. IEEE Published Journal Paper, 2x National Finalist, and leading a 150-member ACM chapter. I don't just learn AI, I build with it.
The tools I reach for to build production AI systems — from model training to deployment.
Languages
AI / LLM / RAG
Computer Vision
ML
Full-Stack & Cloud
Neal Daftary's professional experience: AI Intern at 8xSports (June–September 2025), building a real-time visual search engine with YOLOv8, DINOv2, and FAISS achieving 645ms latency across 301 sports profiles — earned Letter of Recommendation. AI Software Engineering Intern at MZHub Faithtech (October–December 2025), a spiritual technology platform for religious institutions, improving web performance by 57% and SEO by 40% on Azure. Undergraduate Student Researcher at Nirma University under ISRO funding (January 2026–present), building Computer Vision pipelines on Chandrayaan-2 TMC-2 and OHRC lunar imagery for automated crater detection, segmentation, and morphometric analysis.
AI Intern (Remote)
Maintained sub-800ms retrieval latency at 50× scale (15,000 athlete profiles) via a FAISS IVF-PQ/OPQ index with GPU batch encoding and nightly incremental updates. Cut edge-device inference to sub-20ms by fine-tuning YOLOv8s and exporting to ONNX + TensorRT FP16. Improved re-identification robustness under occlusion by fusing DINOv2 ViT-S/14 patch embeddings with LBP texture histograms.
Software Engineering Intern (Remote)
Achieved +57% performance and +40% SEO gains (Core Web Vitals) by deploying an SSR Next.js app to Azure App Service via a GitHub Actions CI/CD pipeline. Integrated Azure Cosmos DB and SendGrid SMTP for scalable data persistence and automated transactional email. Shaped the product roadmap through R&D on agentic AI workflows and CX-automation chatbots.
Open Source Contributor (Remote)
Contributing production code to jax_privacy, Google DeepMind's JAX-based differential privacy library — authoring and merging pull requests that extend its DP feature set, and refactoring the chex dependency to simplify setup for downstream users. Contributions reviewed and approved by the maintainer team.
Neal Daftary's AI and ML projects include: SOLV.ai — AI voice complaint management using ONNX DistilBERT-MNLI and MiniLM-L6; Production RAG Chatbot for IAT Networks using ChromaDB and Groq with 4-layer guardrails; MemoryLens — LLM memory decay benchmark with 5.45× recall improvement; SpectraScan — AI paint defect detection, 4th National at Mitsubishi Electric Cup 2026; Lumin.AI — solar inverter predictive maintenance, HACKaMINeD 2026 national winner; Visual Search Engine at 8xSports at 645ms; CatBoost anomaly detection published in IEEE Sensors Letters.
AI systems, research, and full-stack products shipped in production — across computer vision, LLM infrastructure, and web engineering.
B.Tech in CSE (Artificial Intelligence & Machine Learning)
CGPA: 7.77 | JEE Mains 2024: 97.62 percentileAssociation for Computing Machinery (ACM) ITNU
Leading 150+ member technical community. Launched 'Prompt to Prototype' and mentorship tracks across AI/ML, Data Science, Cybersecurity, and Web Dev.
Got a challenging AI problem, a product to ship, or research to collaborate on? I'm always down to build something exceptional.
