Experience
Production systems across AI, telecom, and web platforms.
Nokia
Software Engineer, AI Platform Engineering · Sunnyvale, CA
Aug 2024 – Present
The AI Platform Engineering team builds internal developer tools that accelerate 5G product development across Nokia's Cloud & Network Services business. My work focuses on applying LLMs and retrieval systems to telecom test automation, observability, and diagnostics.
AI Test Automation Platform
Problem: Telecom PRDs are hundreds of pages of unstructured prose. Manually converting them into executable regression test plans took weeks per release across multiple QA teams.
What I built: A production AI platform that ingests PRDs, extracts testable requirements via LLM-guided parsing, and generates regression test plans grounded in the source document. Used by 4 internal engineering teams.
How it works: FastAPI backend, hybrid retrieval (FAISS IVF + BM25 + reciprocal rank fusion), structured output via function calling, evidence-scoring guardrails to prevent hallucinated requirements.
Python, FastAPI, FAISS, sentence-transformers, Docker, Kubernetes, GitHub Actions
LLM Diagnostics Copilot
Problem: Root-cause analysis of 5G system failures required manual correlation of logs, command outputs, and historical tickets by senior engineers.
What I built: An LLM-powered diagnostic interface that suggests probable root causes, relevant commands to run, and historical similar incidents — with an evidence-citation constraint.
How it works: RAG over internal ticket corpus + command reference docs. Every suggestion must cite a source. React frontend, FastAPI backend.
Python, FastAPI, FAISS, Qdrant, React, Docker
ATS Intelligence (Anomaly Detection for 5G Logs)
Problem: 5G network elements produce 500K+ daily log events; manual anomaly spotting does not scale.
What I built: A real-time anomaly detection engine using statistical baselines and z-score deviation on structured log features, reducing mean-time-to-identify from ~45 min to ~12 min.
How it works: Streaming log ingestion → feature extraction → per-series z-score computation → threshold alerting with severity scoring and LLM-generated natural-language summaries for on-call.
Python, FastAPI, Prometheus, Grafana, PostgreSQL
Technical environment: Python, FastAPI, FAISS, Qdrant, Docker, Kubernetes, PostgreSQL, Redis, GitHub Actions, Linux, gRPC
Rochester Institute of Technology
Software Engineer (ITS Application Development) · Rochester, NY
Jan 2023 – Aug 2024
- Built a search-driven Transfer Credit Equivalency platform using Apache Solr, Node/Express, GraphQL, and React, serving 15K+ student queries per semester.
- Designed and deployed REST + GraphQL APIs with optimized Solr indexing and query pipelines, improving search relevance by 30% (measured via NDCG@10).
- Developed campus eServices web applications (Angular, React, Java, PHP) with SSO auth integration, used by 20K+ students and staff.
- Integrated AWS Lambda, S3, and Cognito for event-driven document processing workflows, handling 8K+ document uploads per month.
- Shipped 12 production features end-to-end across 3 release cycles, from UI to API to data layer to deployment.
Ericsson
Software Engineer, RAN Software (LTE/NR) · Bangalore, India
Feb 2022 – Aug 2022
- Engineered high-throughput backend services in C++ and Python for LTE/NR telecom protocol stacks, reducing packet processing latency by 20% (measured via internal perf benchmarks).
- Designed telemetry pipelines and internal service APIs supporting real-time KPI monitoring for 3GPP L2/L3 layers across 50+ base station configurations.
- Built CI/CD automation with Jenkins and Ansible, cutting release cycles from 4 days to 3 days (25% reduction) and achieving 99.2% deployment success rate.
- Collaborated on fault tolerance and network topology orchestration for distributed RAN systems spanning multi-vendor environments.
“This is where I first saw how much manual effort was going into telecom testing — the seed for my later work on AI-assisted automation at Nokia.”