Shreshth Goyal

Software Engineer @ CureFit

I build production AI systems at CureFit — RAG agents serving 1M+ requests/month, MCP-orchestrated on-call agents that cut resolution turnaround by 98%, and a WhatsApp CRM integration processing 700K+ messages/month. My research on approximate nearest neighbor search (MAVEN) achieved sub-0.5ms query latency on edge devices. I also built a systematic equity trading engine with 43.27% out-of-sample CAGR. LeetCode Knight (1972), dev.to moderator since 2022.

Experience

Software Engineer at CureFit

Jul 2025 - Present | Bengaluru, India | 1 year 2 months

Built ICS support and on-call agents using MCP tooling and sub-agent orchestration on an internal harness, reducing resolution turnaround time by 98% for support and 97% for on-call, while cutting manual escalations by 55%.
Integrated WhatsApp messaging into an internal CRM with callback handling and event publishing; system processes 700K+ messages/month across 105K users with PII-safe audit logging, increasing lead dispositions by 35%.
Replaced a queue-based update mechanism with direct user-attribute updates for real-time trial booking in the CRM, cutting scheduling latency from 10 minutes to near real-time and recovering a 12% degrowth in trial attendance and conversion.
Built a call transcription and LLM summarization pipeline auto-categorizing 4000+ inbound leads/month with pre-dial intent signals, improving follow-up conversion 15-20% and achieving a 10.18% call-to-conversion rate.

Software Engineer Intern at CureFit

Jan 2025 - Jun 2025 | Bengaluru, India | 6 months

Orchestrated a centralized bulk lead-upload pipeline and a contextual inbound callback system, increasing lead dispositions to 94% from 62% and conversions by 13%.
Owned checkout drop-off recovery, an automated system for new and renewal abandonments, recovering 45K+ lost conversions (12% of lost leads), lifting new user conversion rate by 5% and renewal rate by 7%.
Developed an AI-driven vulnerability-resolution workflow that auto-generates draft PRs with CVE context, reducing patch turnaround by 98%, cutting critical and high vulnerabilities 70%, and improving overall turnaround by 80%.
Designed a geographical seller-attribution system enabling IPO-grade financial integrity, preventing large-scale misattribution worth $90K+.

Software Engineer Intern at CureFit

May 2024 - Jul 2024 | Bengaluru, India | 3 months

Developed a RAG-based customer support agent serving 1M+ requests/month using FAISS, cutting support turnaround time from 24 hours to 5 seconds with 75% improved retrieval precision.

Moderator & Top Author at Forem (dev.to)

Apr 2022 - Present | Remote | 4 years 5 months

Published 6 articles on dev.to with 500+ total reactions, covering RAG chatbots, CLI tools, Tree-sitter, and NodeJS authentication.
Managing community interactions and maintaining content quality standards on the Forem platform.