Chris – AI agent development, LLM, LangChain

A legal technology SaaS platform that uses a photorealistic HeyGen AI avatar integrated with a custom LLM backend to deliver secure, cost-efficient, and conversational guidance for real legal workflows, supporting sensitive data and extended user sessions.

• Led and delivered all avatar engineering for a legal-tech SaaS platform, working independently while coordinating with a separate LLM backend team;
• Implemented a photorealistic real-time avatar with low-latency lip-sync using HeyGen, meeting requirements for visual quality and session reliability;
• Architected the avatar system to support extended sessions (1+ hour) with cost efficiency and strong data security constraints;
• Integrated the avatar layer with a custom LLM backend, ensuring seamless handoff between AI responses and real-time rendering;
• Delivered full source code and a production-ready Docker image to support scalable deployment across multiple avatars.

An AI-driven drug test screening pipeline that replaces manual image review using computer vision, OCR, and an LLM layer via AWS Bedrock (Claude), built with HIPAA-compliant data handling and secure cloud infrastructure.

• Engaged as sole AI specialist to own the full technical roadmap, strategy, and delivery for the client's first AI initiative;
• Designed and built an end-to-end automated drug test screening pipeline using OpenCV for image processing, PaddleOCR for text extraction, and AWS Bedrock (Claude) for intelligent result interpretation;
• Replaced a previously manual image review process, enabling screening at scale with significantly reduced human effort and error;
• Developed the system with HIPAA compliance considerations throughout — covering data handling, access controls, and secure cloud infrastructure;
• Operated as a trusted technical advisor, guiding a non-technical client through architecture decisions, risk assessment, and delivery planning.

A digital health platform that combines an AI avatar–led pre-assessment with clinical evaluation to deliver structured ADHD insights, streamlining patient intake and improving diagnostic consistency for healthcare providers.

• Acted as an AI systems consultant for an AI-powered ADHD assessment platform used in clinics and hospitals;
• Designed and implemented a robust Firestore data architecture covering patients, interview transcripts, session backups, and clinician outputs;
• Refactored and stabilised the codebase, resolving schema inconsistencies and aligning frontend, backend, and avatar systems;
• Collaborated with the founder on technical roadmap, data design, risk analysis, and investor readiness;
• Ensured high reliability and data integrity standards suitable for clinical environments.

A seasonal consumer SaaS platform enabling real-time personalized video calls with an interactive Santa avatar, combining low-latency streaming, conversational AI, and a conversion-focused funnel to handle high traffic and drive revenue.

• Conceived, designed, and delivered a seasonal AI product end-to-end, owning product strategy, system architecture, and business direction from idea to launch;
• Hired and led a team of 6, coordinating across engineering, design, and operations throughout development and live season;
• Personally implemented core AI, backend, real-time streaming, and DevOps, building a Dockerised FastAPI backend on AWS EC2 with Nginx, Supabase (auth, database, RLS), and Stripe for payments;
• Integrated real-time video avatars via HeyGen and LiveKit, combined with OpenAI Realtime API for streaming LLM reasoning, low-latency STT, and natural conversational responses;
• Designed a conversion-focused funnel with free preview, IP-based rate limiting, bot protection, and tracking to maximise paid upgrades during high-traffic periods.

An AI-powered coaching platform featuring a conversational avatar that delivers personalized, memory-driven coaching sessions through real-time video interaction, adapting to user goals over time.

• Led end-to-end AI R&D across the coaching and leadership product suite;
• Designed and shipped a production SaaS AI coach with real-time video avatars, voice interaction, and persistent user memory;
• Integrated HeyGen’s real-time avatar API to enable synchronized video and voice for a human-like coaching experience;
• Architected the full-stack system, including LLM, STT, and TTS pipelines for low-latency conversational flow;
• Built and maintained WebSocket infrastructure to support real-time audio and video avatar synchronization;
• Managed cloud deployment on AWS using Nginx and Docker for scalable and reliable delivery;
• Owned the full development lifecycle from architecture decisions to deployment and iteration.

An intelligent NPC system for a multiplayer metaverse platform, using a custom LLaMA 2 model with persistent memory and UE5 MetaHumans to deliver lifelike, real-time voice and text interactions, supported by a configurable web app and API layer.

• Oversaw the design and implementation of an AI-driven NPC system, leading all game-side integrations between AI components and Unreal Engine 5;
• Integrated Whisper STT (locally hosted) and ElevenLabs TTS to enable real-time voice interaction with NPCs;
• Collaborated with the AI team to connect a custom-trained LLaMA 2 model (hosted on Paperspace) into the UE5 environment;
• Implemented dynamic NPC memory for personalised, persistent interactions across multiplayer sessions;
• Worked with UE5 MetaHumans to deliver realistic facial animations, emotions, and lip-sync driven by live AI output;
• Contributed to a scalable API layer enabling integration across multiple game engines and third-party platforms;
• Led agile ceremonies and translated product requirements into actionable engineering epics and stories.