Applied AI specialist and NLP practitioner with a Ph.D. from IIT Bombay and 6+ years building production AI systems in the financial industry. Builds LLM infrastructure at ISS STOXX — covering deployment, observability, and optimization — and serves as the organisation's go-to resource for NLP/LLM understanding and adoption, including a company-wide workshop series for software development teams. Published in ACL, EMNLP, and COLING; brings research-grade rigour to practical, scalable AI solutions.

• Deployed and maintain multiple LLMs on internal GPU infrastructure using vLLM (Gemma3:27B), applying FP8 quantization, kv-cache quantization, multi-GPU tensor parallelism; diagnosed CPU-level AVX2/AVX-512 bottleneck that led to a server hardware upgrade. Manager described this as a "break-through moment for my team and the company."
• Designed and currently delivering company-wide "Introduction to LLM for Software Developers" workshop series (8 sessions planned; 3 delivered as of Mar 2026) using a flipped-classroom format. ~70 participants per live session; additional reach via recordings.
• Grew the ESG news pipeline's business relevance rate from ~2% to ~35% — analysts now need ~3 reads per relevant article, down from ~50. Led a Japanese-language expansion that raised business relevance for Japanese content from ~9% to ~33%. The pipeline handles ~700K articles/day across 10+ languages.
• Built and open-sourced vLLM Metrics Dashboard — a Telegraf → InfluxDB 2.0 → Grafana stack monitoring TTFT, throughput, KV-cache utilisation, HTTP responses, and Python GC across 3 vLLM inference instances.
• Built an AI-driven document processing pipeline for thematic index creation (PoC): PDF → Markdown → LLM summarisation. Business stakeholders validated the output positively.