Haris Ash

AI/ML Engineer building AI agents, automation pipelines, and generative AI applications — production experience spanning RAG systems, multi-agent orchestration (LangGraph, MCP, A2A), async backends, and real-time LLM-powered systems on Vertex AI, Pinecone, and Redis.

Founding AI Engineer at Hike Tech LLC (2025-11 – 2026-05)

• Engineered Hike Study, a multimodal AI tutoring app Python (Quart) async backend — live on the App Store and Google Play Store — scaling to 2,500+ subscribers.
• Cut Time-to-First-Token from 13s to 3–4s (73%) by redesigning the multimodal (text/image/PDF/voice) inference pipeline on Vertex AI with SSE streaming.
• Built a dual-layer RAG pipeline (Pinecone + Redis) separating static syllabus content from personalized user uploads – cached queries under 50ms, token costs down 20–30%.
• Kept the async event loop unblocked under load by offloading PDF processing and background tasks to multithreaded workers.
• Prevented cross-user data leakage by enforcing per-user data isolation across 768-dim embeddings via metadata-filtered retrieval.

ML Engineer Trainee at CDAC (2025-04 – 2025-07)

• Built OpenCV/YOLO computer vision pipelines for context-aware object detection, improving detection accuracy ˜60% and reducing false positives.
• Deployed models into a Flask web app for real-time inference; built scraping and embedding pipelines feeding a semantic cache

AI/ML Engineer Trainee at Rooman Technologies (2024-10 – 2025-04)

• Built LangChain-based agentic pipelines for autonomous reasoning tasks and Python (Pandas/NumPy) pipelines to process large-scale datasets

B.E. in Artificial Intelligence and Machine Learning – PA College of Engineering (2021-12 – 2025-07)