CSE freshman at NIT Calicut with strong foundations in AI/ML, competitive programming, and scalable system design. Building intelligent systems that are reliable, not just accurate. Open-source enthusiast with research interests in model interpretability, AI security, and modern ML paradigms.
Gmail extension with deterministic 3-bullet summarization, strict label categorization via DOM parsing, and a hybrid spam engine combining Gemini inference with weighted heuristics (0–100 score).
GitHub ↗Secure medical timeline platform with voice-to-note capture, Qdrant semantic retrieval (384-dim embeddings), Groq Llama 3.3 70B analysis, and PDF export for doctor-ready reports.
GitHub ↗Web-based genetic risk assessment predicting 10 inherited diseases via weighted clinical scoring. Combines Family (40%), Lifestyle (35%), and Lab Values (25%) with Class I–IV stratification.
GitHub ↗Distributed compute mesh with coordinator–worker architecture for isolated Python job execution across LAN nodes. TypeScript SDK for job submission, worker registration, and credit-based execution.
GitHub ↗Contrastive activation steering without weight updates to control LLM uncertainty at inference time. Layer-wise hidden-state injection with automated evaluation of semantic similarity and uncertainty markers.
GitHub ↗Six ML models implemented end-to-end with real-time interactive demos. Includes from-scratch autograd engine (MicroGrad), MNIST digit classifier with canvas drawing, GPT-2 Transformer modules, and animated forward-pass visualizations.
GitHub ↗Real-time gesture-controlled holographic interaction system. 3D gyroscopic rings, orbit-core demos, and a 2-link robotic arm with inverse kinematics via MediaPipe hand landmark detection.
GitHub ↗Wave-based browser survival shooter with boss progression, laser sweeps, teleport slams, delta-time game loop, real-time particle engine, and adaptive WebAudio — zero external assets.
GitHub ↗Microcontroller-based real-time Wi-Fi latency measurement using packet timing diagnostics. Memory-efficient buffering and serial reporting via lightweight embedded firmware.
I'm Joshua Jacob Thomas — first-year CSE at NIT Calicut, building at the intersection of AI research and systems engineering. Strong foundations in AI/ML paradigms, competitive programming, and scalable system design. I care about making intelligent systems that are reliable — not just accurate.
Open-source enthusiast and active member of CP Hub, Cyber Security Club (AI Exploitation), FOSSCell, GDSC, and DebSoc at NITC. Serving as Placement Representative at the Centre for Career Development.
Interests: mechanistic interpretability, AI security, modern ML, distributed systems, algorithms, and the occasional browser game built from scratch at 2am.