Rashid

Independent AI Consultant

Freelance • Freelance

Jan 2025 - Present • 1 yr 9 mos

Independent AI architecture and delivery for clients in mobility, enterprise software, hospitality, consumer devices and insurance. Selected systems: - Sentinel (highway operators): Architected a five-product edge vision suite (plate reading, safety violations, incident alerts, vehicle and axle classification, traffic counts) running live at expressway and toll sites on modest roadside hardware. YOLO detection, tracking, OCR, OpenVINO optimization, and smoothing for full-frame-rate evidence video. - Aegis (enterprise software): Designed and built an AI governance gateway that masks personal data before any model call, enforces topic policy and approved models in one place, and logs every call, including multi-agent chains, for audit. Demonstrated live to global enterprise clients. - Conductor (restaurant brands): Architected a multi-agent marketing backend, an orchestrator plus 7 specialist agents for content, campaigns, search, social and reviews, with event-driven messaging, replayable failures and per-client cost metering. - Echo (consumer devices): Built the prompt and model architecture for an audio-only Hinglish and English assistant for smart glasses, using a small router model with a larger specialist model across ten domains. - Meridian (insurance planning): Turned a one-client territory and quota planning copilot into a reusable product on a semantic data layer, re-modelled for 10 commercial insurance product lines in 10 working days. - Tideline (insurance): Built an in-platform premium forecasting app with an auto-selected model zoo, driver-based what-if scenarios and AI-written forecast narratives, so client data never leaves the platform.

Senior Data Scientist

AgreeYa Solutions

Feb 2024 - Present • 2 yrs 8 mos

Own the architecture and delivery of enterprise GenAI systems at platform scale. - Designed RAG-based GenAI architectures combining vector databases, documents, PPTX and relational data with prompt orchestration to minimize hallucinations. - Led distributed fine-tuning of transformer LLMs with PEFT (LoRA), PyTorch FSDP, DeepSpeed and HuggingFace Accelerate, achieving 3x faster training. - Built low-latency, high-throughput inference pipelines with vLLM, TensorRT and OpenVINO for real-time enterprise workloads. - Architected event-driven AI workflows for asynchronous execution and independent scaling of GenAI components. - Delivered document intelligence and proposal automation platforms that cut manual effort by 90%+. - Recognized by AgreeYa leadership for architecting an enterprise-scale GenAI backend with 95% automation efficiency.

Data Scientist

ElectrifAi

Feb 2023 - Feb 2024 • 1 yr

Led client engagements to understand business challenges and deliver tailored GenAI solutions. - Built applications on Large Language Models and enhanced open-source LLMs to solve complex business problems. - Fine-tuned open-source LLMs with PEFT (LoRA/adapters) on call-center transcripts to extract contextual conversation features between agents and customers, reducing hallucinations. - Used distributed multi-GPU training via HuggingFace Accelerate to cut training time by 3x while maintaining accuracy. - Developed a Retrieval Augmented Generation (RAG) chatbot that resolved domain-specific hallucination issues. - Automated text analytics on news articles with GenAI to extract weaponry-related events. - Hands-on with LangChain, prompt engineering, prompt formatting and LLM fine-tuning.