I build production AI products end to end: LLM, RAG & agentic systems, on-device computer vision, and scalable Next.js / FastAPI platforms. From fine-tuned models to the last 10% of the UI, I ship things people actually use.
Engineer at the intersection of AI & the web
Hi, I'm Vishwam Shah, an AI engineer and full-stack developer finishing an M.Tech in Artificial Intelligence at PDEU, after a B.E. in Information Technology at GTU. I build production systems that pair real machine learning with clean, scalable web apps.
My work spans LLM, RAG & agentic pipelines, on-device computer vision (YOLO / OCR shipped to edge devices via ONNX), and multi-tenant SaaSon Next.js and GCP. I've shipped platforms powering $50K+ in transactions, automated 40 hours of manual work a week for a client, led small teams, and published BCI research accepted at an IEEE conference.
I care about the whole loop, from training the model to the last 10% of the interface.
The stack behind the work
A focused toolkit across the AI and web spectrum, chosen for shipping real products.
Languages
AI & Agentic AI
RAG & Vector Search
ML & Deep Learning
Computer Vision & Edge
Frontend
Backend & APIs
Data, Cloud & DevOps
Things I've built
A mix of AI systems and full-stack products, each shipped to real users.
Antrenor AI
An AI interviewer platform running real-time voice interviews over LiveKit: it refines your CV for a target role, generates role-specific questions and scores each answer automatically. Backed by a curated bank of real interview questions scraped from 130+ companies across multiple verified sources.
CHATZEN
A multi-tenant conversational AI platform that unifies WhatsApp, Instagram and other channels in one interface, with tenant-specific knowledge grounding and centralized management for every conversation.
MailPilot
A tool-using AI email assistant that takes natural-language instructions to reply, forward and organize mail, classifying each message's intent to cut manual triage down to the messages that actually need a human.
Knowledge Base Portal
A RAG knowledge system using vector databases and LLMs for contextual Q&A and semantic search over large enterprise document sets.
ANPR Mobile Application
A React Native field app that reads vehicle number plates from the live camera feed - a custom YOLO detector plus a fine-tuned OCR model, run fully on-device through ONNX Runtime at 83% accuracy and sub-300ms per frame.
AI Content Generator
A dashboard for structured, SEO-optimized blog creation across multiple formats, with one-click export and multi-site publishing powered by LLM APIs.
Document Intelligence Engine
An AI pipeline for OCR, structured information extraction, automated document classification and multi-document summarization at scale.
Financial Sentiment Platform
Combines FinBERT, custom NLP pipelines and real-time market data feeds to generate predictive sentiment insights on Indian equities.
The road I've travelled
Five roles across AI engineering and full-stack product - training models, shipping platforms and leading small teams.
AI Engineer · HunterRepo
- Owned the ML pipeline for a vehicle-recovery platform end to end: curated a custom dataset, trained and validated a YOLO plate-detection model, and shipped it at 83% accuracy with sub-300ms on-device inference.
- Deployed to edge devices via ONNX Runtime and native Kotlin & Swift modules, removing server round-trips entirely, and built the supporting Python services and operations portal.
- Created a continuous-learning loop with an active-learning gate that auto-confirms high-confidence reads and routes uncertain crops to CVAT for retraining.
- Architected a 12-step repossession state machine with RBAC, real-time SSE notifications, automated invoicing and a full audit log.
Full-Stack AI Engineer · RealtyEaseAI
- Architected and shipped LLM features into a production multi-tenant SaaS on GCP, integrating Groq to automate lead capture, document handling and task tracking - cutting 40 hours of manual work per week.
- Designed and tuned robust prompt pipelines for classification, information extraction and content generation, iterating on accuracy, response quality, latency and token cost across multiple AI providers.
- Built a production RAG pipeline over client documents with intelligent chunking, embeddings and vector search, grounding assistant answers in company-specific data and sharply reducing hallucinations.
- Delivered the platform end to end with RBAC, secure APIs and real-time messaging, supporting $50K+ in transactions, while leading a 2-person team through code reviews, testing and releases.
AI Engineer · Espread Solutions
- Built AI content-generation systems on OpenAI / Gemini / Groq with custom prompt-engineering pipelines.
- Developed an AI keyword generator, SEO automation and structured multi-step content workflows.
- Shipped a guest-posting platform with separate Admin/User dashboards and end-to-end publishing.
Summer Intern · DIAT - DRDO
- Built and evaluated deep-learning sentiment-analysis models for a government defense use case, classifying text as positive, negative or neutral with NLP preprocessing and feature extraction.
- Developed the full pipeline - preprocessing, data preparation, training, hyperparameter tuning, evaluation and real-time inference - benchmarking models on accuracy, precision, recall and F1.
Software Developer Intern · Demaze Technologies
- Built a production-ready RBAC dashboard, secure authentication system and admin panel for VidhyaX, with protected routes, secure session handling and permission-based UI rendering across multiple user roles.
- Developed and integrated RESTful APIs for user, role and administrative operations, keeping frontend and backend in step behind a secure, scalable access layer.
- Improved page performance ~25% through lazy loading and code splitting; shipped 5+ reviewed pull requests.
Published work
EEG-Based Motor Imagery Classification Using Deep Learning Techniques
M. Dsylva, V. Shah, S. Acharya, S. Vankadiya, Dr. S. Satapathy, Dr. N. Singh
Certifications
Talk to my AI assistant
Drop a quick message and it'll guide you through reaching me, then deliver it straight to my inbox.
Type your message and press Enter
Let's build something great
Have a role, a project, or just an idea worth chasing? My inbox is open.
Get in touch
I'm open to full-time roles, freelance work and research collaborations. I usually reply within a day.