Lead Software Engineer – Machine Learning
Freshworks, Chennai, India · Apr 2023 – Present
Freddy AI Insights — Key Co-Architect
- Co-architected a distributed intelligence engine surfacing anomalies, trends, and root cause analysis for enterprise customers, processing ~32M data-generation queries and ~200M ML evaluations daily across tens of thousands of tenants.
- Led AI/ML layer: statistical profiling algorithms (distributional skew, monotonicity detection, longest-contiguous-subsequence analysis, change-point detection) executed as in-memory Spark batch jobs within a strict 1-hour SLA.
- Designed a multi-stage async architecture (SQS + Kafka + Spark Streaming + Spark Batch) achieving 99.99% SLA compliance and ~5% of legacy pipeline cost while processing 10× the data volume.
- Built a config-driven ML pipeline (YAML → dynamic Databricks Spark DAG) allowing new products to onboard new insight types in days.
- Coordinated delivery across 6 cross-functional teams; presented architecture at the Freshworks Architects’ Forum and co-authored the Freshworks Engineering Blog post (March 2026).
Conversational Analytics – Agentic AI
- Built a natural language to chart/graph system for the Freshworks Analytics Platform, serving 17K+ requests across 4K+ accounts in production.
- Designed a fully custom agentic orchestration system (predating LangChain/LangGraph) with custom tool routing, state management, and multi-turn conversation handling.
- Implemented RAG pipelines grounding the system in product-specific data schema and metric definitions.
GenAI Evaluation Framework — Current Initiative
- Architecting a standardised evaluation pipeline for all Copilot GenAI features, unifying the AI Agent Platform with Databricks/MLflow via Arize Phoenix.
- Designed a Medallion data strategy (Bronze → Silver → Gold Delta tables) normalising live traces and offline simulation data for LLM-as-a-Judge evaluation.
- Enabled offline experimentation using curated Golden Datasets against sandboxed test agents.
MLOps & Infrastructure
- Managing AWS infrastructure (VPC, Kubernetes, RDS, SQS, S3, EFS, Lambda) for the Neo Analytics AIML team.
- Optimised pipeline costs by ~24% through Karpenter autoscaling, Nitro instance migration, and VPC endpoint optimisations.
- Managing Databricks workspaces and Unity Catalog across 5 regions.
Senior Software Engineer – Machine Learning
Freshworks, Chennai, India · Oct 2021 – Mar 2023
- Led a team of 4 to design and deliver a multi-product, multi-tenant custom intent detection system handling 1M+ requests/day from 10K+ active bots in production.
- Architected the full E2E system: FastAPI + Celery microservices, real-time inference, and active learning pipeline backed by MongoDB, Redis, S3, and Elasticsearch.
- Used LaBSE multilingual sentence embeddings for cross-language intent detection without per-language model variants.
- Built Freshworks’ first MLOps platform on Databricks covering experiment tracking, feature store, and model registry; integrated KServe for production model serving.
- Coordinated with 8 cross-functional teams for delivery and 3 additional product teams for integration.
Software Engineer – Machine Learning
Freshworks, Chennai, India · Apr 2020 – Sep 2021
- Extended the core bot platform to Freshchat and Freshservice, scaling to 60K+ customers and 200K+ bots.
- Fine-tuned a BERT model for IT Support Bot Service Item Suggester on Freshservice, deployed via TensorFlow Serving.
- Developed a V2 Solution Article Suggester using MUSE embeddings with improved accuracy and cross-language relevance.
Graduate Trainee – Software Engineering
Freshworks, Chennai, India · Jun 2019 – Mar 2020
- Built the core multi-tenant model training and data sync pipeline for Freshdesk bots, serving 42K+ customers and 90K+ bots. Stack: Java, Spring Boot, Kafka, MySQL, Redis.
Education
Bachelor of Engineering – Computer Science Sathyabama Institute of Science & Technology, Chennai · 2015 – 2019