SpeaklineAbout Me
Software Engineer (SRE/AI) at Cisco building production-grade AI/ML systems. I architect multi-agent systems with LangGraph, RAG pipelines, and FastAPI backends that transform manual SRE processes into autonomous workflows—from LLM-powered incident response to agentic Jenkins automation across enterprise systems.
End-to-End ML Pipeline
/* From model development to production deployment */
// init: agent orchestration
AI/LLM Engineering
Multi-Agent & RAG Systems
LangGraph multi-agent systems, RAG pipelines with intent detection, prompt engineering
// stage: api layer
Backend & APIs
Production Services
FastAPI backends with Pydantic validation, async processing, REST APIs
// deploy: infrastructure
Cloud & Infrastructure
IaC & DevOps
Terraform IaC with GitOps, AWS event-driven pipelines, Kubernetes deployments
// output: production ready
Observability
Monitoring & SRE
Splunk/Grafana dashboards, synthetic monitoring, 98% uptime across 20+ DCs
Skills & Technologies
/* Interconnected knowledge graph */
Languages
layer[0]Backend
layer[1]AI/ML/LLM
layer[2]Cloud
layer[3]Tools
layer[4]Certifications
Professional
Experience
From data science internships and entrepreneurship to SRE & AI engineering at Cisco
Featured Projects
A selection of my recent work across machine learning, developer tools, and AI applications
SpeakLine: Voice-to-Code Comments
Production-grade Python package published on PyPI. Voice-driven inline documentation tool supporting 8+ languages (Python, JS, TS, Go, Rust, Java, C#, Ruby). Pluggable transcriber backends (Whisper local, OpenAI API), language-specific AST parsers, silence detection, comprehensive CLI via Typer, and IDE integration for VS Code, Vim, Neovim, Emacs.
Fine-tuned Vision-Language Model for Image Captioning
End-to-end image captioning system using BLIP achieving 0.1755 BLEU-4 and 0.4011 METEOR scores on Flickr8k. Fine-tuned transformer-based VLM on 8,000+ images with mixed-precision training, achieving 97.5x performance improvement.
Natural Language to SQL Generation using Fine-Tuned LLM
Fine-tuned Llama-2-7B for SQL query generation using LoRA, reducing trainable parameters by 99.4%. Achieved 72% exact match accuracy and 85% BLEU score with 4-bit NF4 quantization enabling training on consumer GPUs.
Gemini Pro Multimodal App
End-to-end multimodal AI application leveraging Gemini Pro for text and image analysis, achieving 60 queries per minute.
Text Summarization & Classification
News summarization using TextRank and K-Means Clustering, achieving 85% precision and 82% F1-score on 1,000+ articles.
Hover to explore
Get In Touch
Interested in AI/ML collaboration or have a project in mind? Let's connect!
I'm always open to discussing AI/ML projects, agentic workflows, or opportunities in the machine learning space. Whether you have a question or just want to say hi, I'll do my best to get back to you!