Gaurav Jha
Machine learning engineer with hands-on experience building production-scale AI automation systems
Hire this person
Send a job offer directly to this candidate
About
Machine learning engineer with hands-on experience building production-scale AI automation systems from LLM-based keyword clustering and SEO content pipelines to multi-source brand visibility tracking. Strong foundation in NLP, graph neural networks, and applied research, with a track record of shipping systems that reduce manual effort and API cost.
Experience
AI Automation Intern - NoBroker.com - Bengaluru, India
(2026-02)
- Fine-tuned a BAAI BGE-large sentence embedding model on domain-specific keyword pairs to power semantic clustering (UMAP + HDBSCAN) of large-scale Search Console keyword exports into primary/secondary intent clusters
- Built an LLM-based (Gemini) cluster refinement layer with an 8-rule validation playbook, reducing manual QA effort on mis-grouped keyword clusters
- Designed a cosine-similarity URL-matching system to auto-classify clusters as "update existing page" vs. "new page needed," feeding an automated, fact-grounded content generator
- Orchestrated the full pipeline as a multi-agent LangGraph workflow with human-in-the-loop approval checkpoints via Google Sheets
- Built a parallel, multi-threaded pipeline querying Claude, ChatGPT, Gemini, and SerpApi to track brand mentions, sentiment, and ranking, processing up to 15,000 keywords/day within API rate limits
- Designed a MySQL-backed, deduplicating global cache that reused prior responses across campaigns to significantly cut redundant API costs
- Implemented rule-based search intent classification (Navigational/Informational/Commercial/Transactional) with AI tiebreaking, and brand-targeted sentiment scoring using a local Hugging Face RoBERTa model
- Built a production Reddit monitoring pipeline using Playwright to bypass anti-bot protections, scraping posts and nested comment threads across target subreddits, with a dual-mode (keyword + semantic embedding) classifier for category matching and confidence-gated disambiguation
- Fine-tuned and published a custom sentiment model (RoBERTa) on Hugging Face Hub to score brand sentiment across posts and comments, automating delivery via Google Sheets and HTML email alerts to stakeholders on a daily cron schedule
Research Intern - IIIT Naya Raipur - Raipur, India
(2024-02 - 2024-06)
- Developed a GraphTranslator framework integrating Graph Neural Networks with LLMs, using a trainable attention mechanism to bridge frozen graph encoders and language models for structured reasoning
- Designed a two-stage training pipeline (graph-text alignment, instruction tuning), improving zero-shot node classification on ArXiv datasets and outperforming BERT, RoBERTa, and LLM-only baselines
Education
BTech(Hon's) - Computer Science and Engineering (Data Science) - Chhattisgarh Swami Vivekanand Technical University (2021 - 2025)



