Debashish Thakur

Leading production-grade LLM agent development for B2B automation, building MCPs with structured output pipelines. Shipped Retell voice agents end to end and built BlueBear Channels bridging Slack and Telegram to sandboxed agents.

Architected LLM/VLM analytics framework processing 10K+ workflow outputs, reducing error rates by 18%. Built ComfyUI-based image generation pipeline serving 3K+ daily requests, cutting inference time by 40% and costs by 30%.

Designed and validated a movie revenue forecasting model achieving 87% accuracy, a 12% improvement over baseline. Engineered an LLM-powered text-to-SQL engine and contributed to LSTM-based models for natural language to SQL conversion.

Pursued a Bachelor of Technology in Computer Science and Engineering, graduating in 2023.