Superfloat
↗AI-first datatype and quantization pipelines for hardware acceleration. Optimizing neural networks for efficient power and speed.
- C++
- CUDA
- Python
- Quantization
Alosh
Denny
AI/ML Engineer & Researcher
Building and deploying agentic AI systems and high-throughput inference pipelines.
Scroll
01
About Me
Computer Science graduate with extensive experience in AI/ML engineering, specializing in deep learning, LLM development, and AI safety research.
Proven track record in building and deploying agentic AI systems and high-throughput inference pipelines. I love involving myself in multi-domain and multi-omic research, pushing the boundaries of Generative AI and Edge Computing.
Core Technologies
- PyTorch
- TensorFlow
- CUDA
- LLMs & RAG
- Computer Vision
- AI Safety
- Quantization
- ROS 2
4+
Years Experience
50+
Projects Built
6
Publications
02
Experience
April 2025 – February 2026
Fullstack AI Engineer
Discern Security, Sunnyvale, CA (Remote)
Architected and deployed agentic AI systems for autonomous threat detection and response. Developed "Agentic Studio" for automated remediation and implemented RL algorithms for cybersecurity.
May 2025 – October 2025
Technical AI Safety Research Contributor
Effective Thesis (Remote) / Effective Altruism
Investigated LLM hallucination rates in long-context models. Designed experiments to quantify evidence distortion and analyzed safety implications of deception-like behaviors.
October 2024 – April 2025
Artificial Intelligence Engineer
Eduport, Calicut, India
Led development of multimodal LLMs for Malayalam education. Launched "Adapt AI", driving 50% MAU growth and achieving #1 ranking in educational apps in Kerala.
October 2023 – June 2024
Data Scientist Intern
NATPAC, Ernakulam
Developed ensemble forecasting models (LSTM, RNN, XGBoost) for Cochin Port traffic prediction with 92% accuracy.
February 2022 – Present
Science and Geology Lead
Team Horizon, CUSAT
Led autonomous Martian rover development for the European Rover Challenge. Built autonomous drone systems using MATLAB and Raspberry Pi.
03
Projects
Superfloat GPU
↗RTL design + GDSII layout for a quad-core Q1.15 GPU built for high-throughput inference-only Superfloat workloads.
- Verilog
- OpenROAD
- VLSI
- RTL
FLUX.1 Abliterated
↗Uncensored Flux image generation model achieved through surgical unlearning techniques to remove baked-in safety filters.
- Diffusion
- HuggingFace
- PyTorch
Swaram
↗Advanced Malayalam text-to-speech model based on conditional VAE architecture for natural speech synthesis.
- VAE
- TTS
- Deep Learning
Reverse SynthID
↗Reverse engineering Gemini's SynthID detection
- Reverse-Engineering
- Watermark Detection
- Adversarial Attacks
04
Technical Arsenal
AI & Data Science
PyTorch, TensorFlow, JAX, CUDA, LLMs, Computer Vision, RAG Systems, AI Safety
Development
Python, C/C++, Triton, Verilog, React, Node.js, FastAPI, SQL
Infrastructure
AWS, GCP, Docker, Modal, Pinecone, Qdrant, Git, ROS 2
05
Positions of Responsibility
2022 – Present
Exploration and AI/ML Lead
Team Horizon
Team Horizon is a multi-disciplinary rover building team from CUSAT. Participated in the European Rover Challenge (ERC) in Poland (2023 & 2024), achieving world rank 11th.
2022 – 2024
Secretary
ACES - Association of Computer Engineering Students
Actively organised and hosted hackathons - CodeCrypt. Hosted sessions on Blockchain, AI/ML, and Research.
2021 – 2023
Member
National Service Scheme (NSS)
Engaged in community services, cleanup drives, and led technical bootcamps for social impact.
06
Selected Publications
DDoS Attack Prediction
↗Regularized Hybrid Deep Learning for DDoS Attack Prediction in Software-Defined IoT Networks. Published in IEEE ICPC2T, 2024.
- IEEE
- IoT
- Bi-LSTM
T2A: Text2ASMR Framework
↗Novel Text2ASMR framework for generating Autonomous Sensory Meridian Response Audio. Preprint, 2024.
- Audio Gen
- Diffusion
- Transformers
e-Healthcare Data Management
↗IoT and ML for Information Management: A Smart Healthcare Perspective. Springer Book Chapter, 2024.
- Springer
- IoT
- Healthcare
Cochin Port Forecasting
↗Time Series Prediction of Container Freight Through Cochin Port Using Machine Learning Models. CTSEM, 2025.
- Time Series
- ML
- Port Logistics
Dementia Stage Diagnosis
↗An Ensembled Deep Learning Model to Improve Diagnosis of Dementia Stages from Brain MRI Images. IEEE ICSCC, 2025.
- IEEE
- Deep Learning
- MRI
Cardiology Research AI
AI Research Methodologies in Cardiology. Springer Book Chapter 6, 2026 (In Press).
- Springer
- In Press
- Cardiology
07
Latest Blogs
Uncensoring Flux.1 Dev: Abliteration
↗An exploration of abliteration and orthogonalization to remove LLM built-in refusal mechanisms.
- AI Safety
- LLMs
- Flux
Unlearning Flux.1 Dev: Abliteration v2
↗Advanced surgical unlearning techniques for Flux.1 Dev, refining the abliteration methodology.
- Diffusion
- Fine-tuning
- Safety
The Ultimate Cookbook: Uncensoring GPT-OSS
↗A deep dive into uncensoring the GPT-OSS-20B model using residual stream ablation techniques.
- GPT-OSS
- Abliteration
- Open Source
The Complete Dummies Guide to LLMs
↗A comprehensive walkthrough of Large Language Models, from transformer blocks to attention mechanisms.
- NLP
- Transformers
- ML Basics
NVIDIA AI Summit 2024: India's AI Future
↗Insights from the Mumbai summit featuring Jensen Huang and Mukesh Ambani on India's technological leap.
- NVIDIA
- Robotics
- AI Ecosystem
Huddle01 SDK: Web3 Communication
↗A developer's guide to building decentralized real-time communication apps with Huddle01.
- Web3
- dRTC
- React
9/10 parents
recommend
Open for collaborations, hackathons, and coffee chats.