SANYAM WADHWA
Building intelligent systems at the intersection of research and production. Agentic AI, LLM optimization, multi-agent architectures, shipped and working in production.
I build things that are
actually useful.
I'm a Computer Engineering student at Thapar Institute, building at the intersection of AI research and production systems. My work spans agentic AI architectures, LLM optimization, and full-stack engineering, always aimed at making intelligent tools that researchers and developers actually want to use.
I work across the full AI engineering stack: from multi-agent systems and RAG pipelines to Chrome extensions with on-device ML, CLI tools that profile hardware and recommend optimal models, and platforms that turn dense academic papers into actionable insights. If it can be made smarter or more autonomous, I'm interested in it.
Previously, I led technical initiatives and event coordination at MLSC Thapar. Outside of building, I explore cutting-edge papers in NLP, multi-agent systems, and LLM efficiency.
The stack behind the systems.
41+ technologies across 6 domains. Hover any row to pause.
Hover any row to pause · Each row scrolls continuously
Things I've built
and shipped.
KONTA
Context-Aware Browsing Extension
Chrome extension for personal knowledge management that captures browsing sessions and clusters related webpages using a session detection and project inference pipeline.
The Problem
Browsers leave no trace. When a tab closes, research disappears. KONTA automatically captures every session and builds a searchable knowledge graph, turning your browsing into a second brain.
How It Works
Key Features
- Automatic browsing session capture with ML-based project inference across open tabs
- Multi-layer search: keyword (BM25) → semantic embeddings (Transformers.js) → ML re-ranking
- Interactive knowledge graph with Louvain community detection across 8+ topic clusters
- Focus mode with context-aware reminders to revisit related browsing sessions
- Fully on-device ML inference via Transformers.js: zero data leaves the browser
Technical Highlights
- Built multi-layer retrieval combining keyword search, semantic embeddings, and ML-based ranking for context-aware discovery
- Developed interactive knowledge graph using Louvain Algorithm to visualize relationships between webpages and projects
Metrics
Tech Stack
KONTA
Context-Aware Browsing Extension
Chrome extension for personal knowledge management that captures browsing sessions and clusters related webpages using a session detection and project inference pipeline.
The Problem
Browsers leave no trace. When a tab closes, research disappears. KONTA automatically captures every session and builds a searchable knowledge graph, turning your browsing into a second brain.
How It Works
Key Features
- Automatic browsing session capture with ML-based project inference across open tabs
- Multi-layer search: keyword (BM25) → semantic embeddings (Transformers.js) → ML re-ranking
- Interactive knowledge graph with Louvain community detection across 8+ topic clusters
- Focus mode with context-aware reminders to revisit related browsing sessions
- Fully on-device ML inference via Transformers.js: zero data leaves the browser
Technical Highlights
- Built multi-layer retrieval combining keyword search, semantic embeddings, and ML-based ranking for context-aware discovery
- Developed interactive knowledge graph using Louvain Algorithm to visualize relationships between webpages and projects
Metrics
Tech Stack
Where I've worked
and shipped.
- Built hybrid agentic mental-health chatbot with modular agents (safety, emotion, sentiment, trend), orchestrated via a Controller LLM with deterministic routing and guardrail enforcement.
- Implemented responsible AI mechanisms including safety filters, risk-aware escalation, and prompt constraints, improving reliability, ethical compliance, and response stability under sensitive scenarios.
- Engineered a full-stack Thapar Datesheet Generator using Flask, MySQL, and Jinja, integrating multithreading to automate end-to-end schedule generation, reducing processing time by 3x vs legacy workflows.
- Incorporated features like subject swapping, room swapping, and real-time resolution of room, faculty, and seating constraints.
Recognition
that followed.
Samsung PRISM Hackathon
Won Samsung's PRISM hackathon for building KONTA toa Chrome extension for personal knowledge management using semantic embeddings and the Louvain community detection algorithm.
Robowars, Saturnalia
Won the Robowars competition at Saturnalia, Thapar Institute of Engineering and Technology.
Let's build
something.
Open to internships, research collaborations, and interesting side projects. Drop a message and I'll reply within 24 hours.
sanyamwadhwa.in@gmail.com