ML Platform Engineer & Tech Lead @ Abacus.AI
I build the reliable platform that AI agents actually run on — not just demos.
I'm Tech Lead for Abacus.AI's Enterprise team, where I own the connector and AI-agent tooling platform serving 1,640 organizations and 46,000+ users who've ingested 900+ TB of data through it. I work end to end — React frontends, high-throughput data pipelines, and the LLM serving, RAG, and fine-tuning underneath.
- 🧠 AI layer: production LLM/agent systems, permission-aware RAG, multimodal page-level retrieval, agent-tool dispatch, LangGraph orchestration, LoRA fine-tuning.
- 🏗️ Platform layer: multi-tenant OAuth, RBAC, vector stores, async/streaming pipelines, retry & rate-limit policies, and the cost/latency tuning that keeps it affordable.
- ⚡ I like turning ambiguous requirements into something shipped — idea to production in ~2 weeks when it counts.
- 🎓 B.Tech, Computer Science — IIT Palakkad.
- 📫 Reach me at raghunathanyogesh@gmail.com
AI / ML
Backend & Data
Frontend
Infra & Ops
- 🔌 A connector ecosystem integrating 25+ enterprise apps & databases (Microsoft 365, Google Workspace, Databricks, Snowflake).
- 🔒 Permission-aware RAG that enforces strict RBAC at query time, so enterprises can trust what their agents retrieve.
- ⚡ 5× faster Microsoft Teams ingestion and 8× faster database exports (validated on a 30M-row Databricks export).
- 🪙 A latency/cost fix on a production LLM pipeline serving 1,000+ enterprise chatbots — input cut ~6×, ~30s misrouting penalty removed.
- 💸 ≈45% (≈$11.7K/month) in AWS storage savings via automated cleanup and dataset-version limits.
Currently building enterprise AI agents at scale — and always happy to talk LLM platforms, RAG, and the reliability layer underneath.