Harsh Verma@vharsh2615709
4
Events attended
2
Submissions made
India
1 year of experience
About me
AI/ML Engineer & Data Science Practitioner | Hackathon Project Lead I am an AI/ML developer and hackathon project leader passionate about architecting scalable solutions using Generative AI, machine learning, and cloud technologies. Currently advancing my expertise through a Master’s in Data Science at Deakin University and having completed a PGP in Artificial Intelligence & Machine Learning from UT Austin, I bridge the gap between deep academic theory and rapid, hands-on execution. As the project lead for the Sūtradhāra team in the Gen AI Academy APAC, I specialize in driving cross-functional initiatives from conceptualization to deployment. My applied experience as an independent contractor and developer spans workflow automation, algorithmic trading model development, and building robust, submission-ready AI applications. I thrive in high-stakes environments, combining technical acumen with strategic leadership to deliver tools that perfectly balance cutting-edge innovation with data privacy and compliance. Core Competencies & Keywords: Generative AI (LLMs, RAG, Agentic AI), Machine Learning, Data Science, Cloud Computing, Algorithmic Trading, Workflow Optimization, Cross-Functional Leadership, Rapid Prototyping, Technical Communication, Product Strategy.
🤝 Top Collaborators
.png&w=640&q=75)
🤓 Latest Submissions
SovereignSwarm: Auto Multi-Tone Video Captioner
SovereignSwarm is an advanced, high-throughput video intelligence application designed to automate the process of localized multi-tone video captioning. Built purely on the fast Fireworks AI inference ecosystem, the system acts as a seamless bridge between computer vision and generative linguistics. The pipeline utilizes OpenCV to dynamically extract key chronological frames from any short video under 30 seconds. These frames are processed through a cutting-edge Vision-Language Model to generate a raw, highly factual description of the scene. The architecture then leverages an asynchronous text-generation layer to rewrite the core description into four completely distinct narrative styles: Formal, Sarcastic, Humorous Tech, and Everyday Humor. The final payload is formatted and returned as a strictly validated, production-ready JSON schema. This ensures seamless integration into enterprise content management workflows, video editing suites, or automated social media pipelines without requiring heavy local infrastructure or complex cloud configuration.
13 Jul 2026
Nadi AI: Autonomous Multi-Agent Response Pipeline
Our project introduces a fully automated, multi-agent incident response system designed to reduce software downtime from hours to seconds. By leveraging the Band SDK for seamless agent-to-agent communication, the system operates as a continuous pipeline that detects, diagnoses, and resolves runtime errors without manual intervention. The Architecture: Agent 1: The Sentry (Diagnostics & Triaging): Acting as the first line of defense, this agent intercepts raw, unstructured crash logs. Using LangChain and Qwen2.5-Coder, it deterministically parses messy logs into strict, Pydantic-validated JSON payloads (extracting error type, severity, and stack traces) and securely publishes them to a designated Band Room. Agent 2: The Patch Engineer (Resolution): Operating asynchronously, this agent monitors the Band Room for new Sentry reports. Upon receiving a structured error payload, it utilizes Mistral (via the Featherless AI API) to analyze the fault and generate a targeted, executable code patch. Agent 3: Command Center & Reviewer (Oversight): The entire ecosystem is visualized through a centralized Streamlit dashboard. Before any code is pushed, a CrewAI-powered Reviewer Agent evaluates the Patch Engineer's proposed fix for security and structural integrity, ensuring human-in-the-loop visibility and control. Key Technologies: Band SDK, LangChain, Qwen2.5-Coder, Mistral (Featherless AI), CrewAI, Streamlit, and Pydantic. Impact: By combining structured LLM outputs with multi-agent networking, we have created a scalable, intelligent DevOps automation tool that treats runtime errors not as blockers, but as instantly solvable data events.
19 Jun 2026
