Headline

AI @Info Edge | Cohere Labs Research ’26 | Open to AI & FDE roles

About

Applied AI Engineer with experience building and deploying AI systems, agentic workflows, and production-focused AI applications. Previously worked on AI products, automation platforms, and multi-LLM systems at Paytm, taking solutions from concept to deployment. Have also won multiple hackathons and have research work accepted at four different conferences. Experienced in Python, FastAPI, model serving, LLM deployment, RAG systems, and AI workflow automation. Currently working at Info Edge on deploying and fine-tuning local language models, reducing dependency on external AI providers while improving cost efficiency and throughput. Passionate about building reliable AI systems that solve real business problems and scale from prototype to production.

Location

Noida

AI stack

antigravity

claude

codex

hugging face

ollama

openclaw

Work experience

AI Engineer at Info Edge India Ltd, May 2026 - Present · 6 mos, Noida, Uttar Pradesh, India, I build LLM systems end to end, from data pipelines to inference, and I've shipped three of them solo at Naukri. * Config Engine: Career page onboarding used to take 30 min per config, manually. I automated it — a Playwright interceptor captures the page's real DOM/network traffic even on lazy-loaded SPAs, an LLM generates the config once, and a validation loop catches bad output and retries before anything ships. Cut setup time by 96%, running on close to zero API cost. * Taxonomy Classifier: Had 31k+ messy skill strings to standardize — duplicates, typos, ambiguous acronyms. Calling an LLM on all 31k wasn't viable, so I clustered similar skills first (cuts the problem down ~80%), ran the LLM once per cluster, then built a fallback chain- exact match, fuzzy match, vector similarity to map anything the LLM missed without calling it again. Ended up at 95% accuracy with 80% fewer API calls. * Job Formatter: Resumes come in messy and inconsistent. Built a two-stage pipeline — deterministic cleanup first, then semantic parsing with SQLite tracking state so it can resume mid-failure instead of reprocessing everything from scratch.

Product Management at Times Internet, Mar 2026 - Apr 2026 · 2 mos, Noida, Uttar Pradesh, India, Worked with the ET team on product initiatives for Soonicorns 2026, contributing to feature development, user experience improvements, and end to end product workflows.

Software Engineer - AI at Paytm, Jun 2025 - Dec 2025 · 7 mos, Noida, Worked in the Paytm Payments vertical within an internal team focused on cost optimization. > Took complete ownership of a product named Mini.Al, leading it from Ideation to Deployment. (a multi-LLM orchestration platform with custom agent deployment, enabling users to leverage multiple Al models in a single workflow.) > Built multiple internal automation POCs that freed the bandwidth of ~3-5 team members across operations. > Collaborated with two other interns on a large-scale migration, achieving ~92% cost reduction.

Summer Intern at Future Crime Research Foundation (FCRF), Jun 2024 - Aug 2024 · 3 mos, Greater Noida, Worked with the analytics team to understand forensic tools, OSINT techniques, and digital evidence analysis while gaining exposure to cyber incident analysis, AI in Cybersecurity and forensic documentation.

Education

Bachelor of Technology - BTech, Computer Science, Galgotias university

Public links

github: https://github.com/nipunraj96

medium: https://medium.com/@nipunraj2004

nipun.framer.website: https://nipun.framer.website/