Agents are entering enterprise apps
Gartner predicts task-specific AI agents will be integrated into 40% of enterprise applications by the end of 2026, up from less than 5% in 2025.
Read Gartner signalSoftware Engineer · Scalable Systems, Test Automation, AI-Assisted Development
Scalable software
with AI systems
judgment built in.
I architect, design, implement, test, and deliver scalable products across Python, JavaScript/TypeScript, SQL, and C++ while using AI coding assistants responsibly and keeping quality, telemetry, and maintainability visible.
Role target
Salesforce-ready software engineering profile
Available for new-grad Software Engineer roles across cloud platforms, AI systems, test automation, and scalable product engineering from May 2026.
900
Unit and functional checks across LLM Forge pipeline stages
4x
H100 SLURM cluster workflow used for scalable AI jobs
80%
Manual processing time reduced through Python extraction service
6+
Software and AI systems spanning SaaS, edge, CUDA, and data products
50+
Research citations across AI, ML, NLP, security
Profile
Built for the space between software engineering and AI systems.
I am targeting Software Engineer roles where architecture, implementation, testing, telemetry, and AI-assisted development all matter. My Business Systems Analyst – AI Automation foundation helps me understand product needs, but my recent work is deeply software-driven.
As a Graduate Research Assistant at George Mason University's Costello College of Business, I architected and built LLM Forge, a configuration-driven 12-stage automation pipeline in object-oriented Python that lets researchers launch jobs from a single config file. It reduced time-to-experiment by over 50% across a shared 4x H100 SLURM cluster.
I designed an automated test framework with roughly 900 unit and functional checks using Pydantic v2 schema validation at every pipeline stage. The goal was exactly what strong engineering teams value: catch configuration errors before runtime, protect shared compute, improve code coverage, and reduce support requests.
I also implemented a Python extraction service using Claude API that converted 500+ multi-page legal PDFs into validated JSON, reducing manual processing time by roughly 80%. Across that work I used Claude Code, GitHub Copilot, and Cursor, but critically reviewed and refactored generated code before merge so the codebase stayed maintainable.
- Architecture firstDefine service boundaries, data contracts, and failure modes before adding features.
- Quality automationUse unit, functional, schema, and telemetry checks to keep systems stable.
- AI with reviewGuide coding agents, review generated artifacts, and keep ownership of the final code.
AI boom hub
Market-aligned, not trend-chasing.
Enterprise AI is moving from loose chatbots to task-specific agents, workflow automation, AI-assisted SaaS, and governed process redesign. This portfolio is written like case-study proof because that is what hiring teams need to evaluate.
SaaS is becoming AI-assisted
Deloitte's 2026 outlook frames SaaS as moving toward more intelligent, adaptive, agent-enabled workflows across enterprise applications.
Read Deloitte outlookImpact matters more than demos
The strongest AI automation portfolios explain the operational problem, implementation path, controls, and measurable change. That is the structure used here.
Read portfolio guidanceSalesforce fit
Software engineering proof for trusted cloud products.
Salesforce needs engineers who can architect, design, implement, test, and deliver scalable products while working with product managers, UX, performance engineers, and AI tooling. This is the engineering signal I want recruiters to find quickly.
Architect, design, implement, test, and deliver scalable products
Built LLM Forge as a 12-stage object-oriented Python pipeline with explicit service boundaries, configuration contracts, shared-cluster execution, and validation gates before expensive runs.
Test automation and release quality
Designed roughly 900 unit and functional checks across schema validation, pipeline stages, and runtime configuration, improving confidence before jobs reached shared H100 compute.
AI-assisted development with code ownership
Used Claude Code, GitHub Copilot, Cursor, Gemini and other agentic tools to accelerate development while critically reviewing generated code, refactoring weak artifacts, and keeping maintainability high.
Operational telemetry and metrics
Instrumented pipeline telemetry and performance metrics to surface low-confidence outputs, failed runs, runtime risk, and quality-control issues before they became support problems.
Cloud and data product foundation
Worked across Python, Java, JavaScript/TypeScript, SQL, C++, HTML, PostgreSQL, Docker, AWS, FastAPI, Linux, SLURM, REST APIs, and multi-tenant product patterns.
Team delivery habits
Translated stakeholder requirements into technical specs, ran weekly reviews, documented decisions, and delivered measurable improvements that product and research users could adopt.
Project atlas
What we built, how we built it, and the impact created.
These projects are structured as product case studies: every one connects a business workflow to frontend, backend, APIs, AI routing, dashboards, settings, and deployment.
Multi-tenant SaaS + payments architecture
SmartRemit — Multi-Tenant Payments Platform
A TypeScript/NestJS and Next.js payments platform design connecting licensed money-service partners to end users through a conversational bot.
What we did
Architected a multi-tenant backend with partner abstraction, end-user workflows, transaction services, and a product surface designed for corridor expansion.
How we did it
Used object-oriented NestJS services, a Next.js frontend, PostgreSQL data modeling, idempotent transaction handling, and clear service boundaries.
Impact created
Showed scalable product thinking: new partners and corridors can be added without destabilizing core payment flows or tenant-specific behavior.
Edge AI + Linux runtime stability
JetBot — Edge AI Agent on Jetson
A Telegram-controlled AI agent running on Jetson Orin Nano with Python, SQLite/FTS5, resource isolation, and human-approval gates.
What we did
Designed an edge AI agent that can search local context, respond through Telegram, and protect destructive actions behind explicit human review.
How we did it
Implemented Python services, SQLite/FTS5 retrieval, Linux systemd deployment, and cgroups v2 resource controls for constrained edge compute.
Impact created
Demonstrated stable AI runtime engineering where inference, search, and automation must share limited CPU, memory, and device resources safely.
C++ systems + performance profiling
CUDA Matrix-Multiplication Kernel
A tiled matrix-multiplication kernel in C++/CUDA focused on memory behavior, shared-memory usage, and performance tradeoffs.
What we did
Implemented a tiled matrix multiplication kernel and benchmarked throughput against the cuBLAS baseline to understand performance ceilings.
How we did it
Tuned shared-memory usage, thread-block sizing, arithmetic intensity, memory access patterns, and occupancy while profiling each iteration.
Impact created
Added low-level engineering evidence for performance-sensitive platforms where implementation details directly affect user-visible scalability.
Request intake + Jira automation
AI Business Request Intake & Jira Automation System
A multi-tenant SaaS product for turning vague stakeholder asks into structured requirements and delivery-ready Jira work.
What we did
Created a self-service intake workspace where teams can submit business requests, generate BRD/FRD-style artifacts, route approvals, and prepare Jira-ready ticket payloads.
How we did it
Mapped request patterns, built tenant auth, client workspaces, automation packs, Jira credential flow, signed webhooks, analytics, and AI research prompts.
Impact created
Modeled roughly 70% faster clarification on sample requests while preserving human review before work reaches engineering.
Automation ROI + process optimization
AI Automation ROI & Process-Optimization Platform
A decision platform for evaluating which manual workflows should be automated first and why.
What we did
Built a workflow portfolio where users can enter process volume, cycle time, errors, cost, integration complexity, and automation readiness.
How we did it
Implemented a six-dimension scoring model, portfolio analytics, workflow detail pages, webhook subscriptions, settings, and tenant-scoped data APIs.
Impact created
Modeled $120K+ annualized savings potential on sample data and turned automation ideas into leadership-ready business cases.
Healthcare workflow + EMR extraction
Healthcare Workflow Automation & EMR Data-Extraction System
A healthcare operations workspace that extracts, validates, and routes synthetic EMR-style records safely.
What we did
Created a clinical queue, intake workflow, record detail view, exception routing, analytics, settings, and research page using synthetic healthcare records.
How we did it
Used privacy-conscious boundaries, tenant auth, server-side validation, clinical event routes, webhook ingestion, and aggregate-only AI research context.
Impact created
Modeled roughly 65% faster extraction and 94% exception detection on synthetic data without using real PHI.
Product master data + quality control
Product Master-Data Automation & Data-Quality Control System
A catalog quality-control SaaS for SKU validation, duplicate detection, approval status, and stewardship workflows.
What we did
Built a merchant workspace for importing product CSVs, validating rule failures, finding duplicates, scoring quality, and managing stewardship stages.
How we did it
Designed quality rules, batch APIs, dashboard analytics, batch detail updates, merchant settings, webhook subscriptions, and catalog event ingestion.
Impact created
Modeled 85% duplicate SKU reduction and item setup improvement from days to hours on a 200+ SKU sample catalog.
Revenue operations + CRM risk scoring
AI-Powered Revenue-Operations Automation Dashboard
A RevOps command center for cleaning CRM exports, scoring pipeline risk, and producing manager action queues.
What we did
Created a revenue workspace with CRM import, forecast dashboard, import lifecycle stages, risk detail pages, settings, webhooks, and AI research.
How we did it
Scored missing next steps, stale activity, close pressure, material deal size, and health signals through server-side APIs and tenant analytics.
Impact created
Modeled weekly reporting reduction from 4 hours to under 15 minutes and surfaced 23 at-risk relationships in sample data.
LLM operations + research enablement
LLM Forge: Domain-Specific LLM Fine-Tuning Platform
A YAML-driven fine-tuning operations pipeline that helps researchers launch repeatable LLM experiments safely.
What we did
Built a 12-stage pipeline for dataset prep, configuration validation, model training setup, evaluation, and experiment tracking.
How we did it
Used Hugging Face, PEFT/LoRA, Pydantic v2 schemas, config-first orchestration, and automated tests to prevent runtime failure.
Impact created
Cut time-to-experiment by over 50% and made shared GPU research workflows easier for non-technical collaborators to use.
Research system
Experience
Where engineering quality became real.
My experience connects professional coding, object-oriented design, test automation, LLM-assisted development, telemetry, stakeholder reviews, and measurable delivery.
- Architected and built LLM Forge, a configuration-driven 12-stage automation pipeline in object-oriented Python that lets researchers launch jobs from a single config file, cutting time-to-experiment by over 50% across a shared 4x H100 SLURM cluster.
- Designed an automated test framework with roughly 900 unit and functional checks using Pydantic v2 schema validation at every pipeline stage, catching configuration errors before runtime.
- Implemented and delivered a Python extraction service using Claude API that converted 500+ multi-page legal PDFs into validated structured JSON and reduced manual processing time by roughly 80%.
- Used Claude Code, GitHub Copilot, and Cursor to accelerate delivery while critically reviewing and refactoring AI-generated code before merge.
- Instrumented pipeline telemetry and performance metrics, surfacing low-confidence outputs and run failures to improve operational visibility and quality.
- Led crews of 5-15 during high-stakes arena setups, enforcing SOPs and reallocating staff in real time to remove scheduling gaps.
- Trained new team members on safety and equipment procedures, reducing setup errors by over 30% and improving time-to-productivity.
- Coordinated event managers, facilities, and external vendors, keeping status updates and incident response aligned.
- Built an end-to-end machine-learning system in Python combining four classifiers with ensemble methods, reaching 93% accuracy and 92% F1 validated through t-test and ANOVA.
- Engineered an Ethereum land-registration prototype using object-oriented Solidity smart contracts and SHA-256 hashing for tamper-resistant transfers across 200 transactions on a 12-node network.
- Evaluated competing models and integrated the higher-performing CatBoost model with 89% accuracy and 91% precision into a web-based analytics dashboard.
Skills
Software engineering capability architecture.
I combine computer science fundamentals, production-minded coding habits, AI-assisted development, and the communication needed to ship features with teams.
Languages and OOP
Java, JavaScript/TypeScript, SQL, C++, HTML, Python, object-oriented design, design patterns, service architecture, REST APIs, and maintainable module boundaries.
Testing and quality automation
Unit and functional testing, automated test frameworks, Pydantic v2 schema validation, code coverage, test strategy, quality gates, and release-readiness checks.
Cloud and product engineering
PostgreSQL, Docker, AWS, FastAPI, NestJS, Next.js, Linux/SLURM, REST APIs, multi-tenant SaaS patterns, dashboards, authentication, tenant models, and webhooks.
AI coding assistants
Claude Code, GitHub Copilot, Cursor, Gemini, prompt engineering, agent-guided implementation, generated-code review, refactoring, and maintainability checks.
Telemetry and operational excellence
Pipeline telemetry, performance metrics, run-failure surfacing, low-confidence output routing, operational dashboards, stakeholder reviews, and support-request reduction.
Credentials
AWS Certified AI Practitioner, NVIDIA-Certified Associate: Generative AI LLMs, Claude Code in Action, Red Hat certification.
Research and publications
Academic rigor behind the automation mindset.
My research background gives me a measurement-first approach: define the process, gather evidence, validate the model, then communicate what changed.
Assessing the Effectiveness of Artificial Intelligence Techniques in Mitigating Cyber Security Risks
Surveyed 468 IT professionals and used SEM/CFA analysis to identify significant drivers in AI-assisted cybersecurity strategy.
Enhancing Hairfall Prediction: A Comparative Analysis of Individual Algorithms and An Ensemble Method
Built and evaluated ensemble ML pipelines, achieving 93% accuracy and 92% F1 across validated model comparisons.
Comparative Analysis of Psychological Stress Detection: ANN and CatBoost
Compared AI models for stress detection and recommended CatBoost based on accuracy, precision, and deployment suitability.
Blockchain, speech quality, and applied AI systems
Co-authored research across Ethereum land registration, speech quality assessment in Indian languages, and intelligent vehicle fuel management.
Education
Computer science foundation.
Formal CS training plus applied research gives me the range to talk to business stakeholders and technical teams in the same project.
George Mason University
Master's degree, Computer Science | 2024 - 2026
KL University
Bachelor of Technology, Computer Science & Engineering | May 2020 - May 2024
Ready for the next system
I am looking for engineering teams building trusted, scalable cloud products.
Best-fit roles: Software Engineer, Salesforce new-grad Software Engineer, Cloud Platform Engineer, AI Systems Engineer, Test Automation Engineer, or product engineering roles where implementation quality matters.
Contact
Let us talk about software engineering work.
If your team needs someone who can learn a large codebase, build features, write tests, use AI tools responsibly, and communicate engineering impact clearly, I would be glad to connect.
Best for direct role, project, and collaboration conversations.
nchennu@gmu.edu
Connect with my full profile, experience, certifications, and research background.
linkedin.com/in/naga-venkata-sai-chennu
GitHub
View code, deployments, portfolio experiments, and implementation work.
github.com/Nagavenkatasai7
Paper Decode
Explore my company/project surface for AI-assisted research and document understanding.
paper-decode.lovable.app