Idris Kulubi@idris_kulubi301
1
Events attended
1
Submissions made
Kenya
5+ years of experience
About me
I'm a 21-year-old Full Stack Software Engineer and athlete with a passion for building impactful, user-centric solutions. My journey has taken me from developing an AI platform for business analysis at KCIC Consulting to creating an all-in-one Event Management System for Nation Media Group that drove significant cost savings and provided crucial data insights.Developed software solutions at Impacting Africa Consulting . I also played a key role in developing application portals for GCA and Youth Adapt with AfDB, streamlining critical climate and development processes. Beyond professional roles, I'm the proud founder of StrathSpace, a social app for university students that boasts over 9,000 users and was even featured on NTV as 'Startup of the Week.' My expertise spans Implementing Artificial Inteligence solutions and Computer programming in Go ,Rust Python, JavaScript, React, Next.js, TypeScript, Node.js, HTML, and CSS, alongside a strong understanding of AI/ML concepts. I thrive on solving complex problems, delivering high-quality user experiences, and continuously pushing the boundaries of what technology can achieve. I'm always eager to connect with innovators and explore new opportunities to build meaningful solutions.
🤓 Latest Submissions
ContinueML
The Problem Generative AI tools (Midjourney, Runway, Pika) are stateless. They have no memory between generations, making it impossible to maintain consistent characters for serialized content like TV series, games, or brand campaigns. Professional creators cite "consistency" as the #1 barrier to adoption. The Solution ContinuuML is the infrastructure layer for AI creativity. We provide a persistent memory engine that stores, retrieves, and validates creative entities to ensure visual and narrative consistency. Technical Architecture Memory Engine: Stores "memory signatures" of characters using multimodal embeddings. We combine visual features (via CLIP, 512-dim) and semantic traits (via OpenAI, 1536-dim) into a Pinecone vector database for robust retrieval. Prompt Enhancement: Uses NLP (spaCy) to detect entity mentions in user prompts. It queries the vector DB and automatically injects specific visual attributes into the prompt before sending it to the generation API. Consistency Analyzer: Validates generated content by extracting embeddings from the output and comparing them to reference memory via cosine similarity algorithms. Branch System: Implements Git-like version control for creative worlds, allowing users to branch storylines without affecting the canonical state. Validation Validated via the Polpharma Challenge, creating commercial AI characters with 93.3% consistency across 48 seconds of video. We reduced production time from weeks to 3 days and costs from $5,000 to $200. ContinuuML solves the consistency problem, turning AI from a toy into a professional production tool
19 Nov 2025
