# Sameer Khan

> Forward Deployed / AI Engineer with founder experience taking ambiguous customer and product problems from discovery through production. Built AI-commerce, bookkeeping, and venture-diligence systems across TypeScript/Next.js, Node.js, Python, LLM tool calling, and AWS; strongest in fast scoping, hands-on implementation, and translating technical systems into measurable business outcomes.

Based in San Francisco, CA. This file follows https://llmstxt.org/.

## Experience

- **Co-Founder & Founding Engineer at agentShop** (Nov 2025 - Aug 2026, San Francisco, CA):
  - Co-founded agentShop, a commerce-intelligence platform for AI shopping; shipped Shopify, WordPress/WooCommerce, and Wix integrations plus an analytics dashboard for tracking AI-driven sales and visibility across ChatGPT, Gemini, and Perplexity.
  - Took the Shopify product to the App Store, launched February 4, 2026, with AI PDP optimization, automated SEO, revenue attribution, AI revenue audits, and AI-shopping recommendation tracking.
  - Shipped product capabilities for share-of-recommendation tracking, product-data enrichment, AI experimentation, competitor visibility, and linking AI exposure to downstream conversions.
- **Portfolio Technology Analyst at MARL Accelerator** (Jan 2025 - Jul 2025, Remote):
  - Conducted technical and strategic diligence on 500+ early-stage startup applications, owning the process from first call through final technical assessment.
  - Built and deployed an AI-powered Analyst Agent that automated initial diligence and generated investment memos aligned with MARL Fund III's thesis, reducing analyst hours per deal by 70%+.
  - Evaluated technical feasibility, architecture, team execution, and AI/product defensibility from incomplete data and unfamiliar codebases under time pressure.
- **Software Developer, AI Bookkeeping at PomJuice** (Mar 2025 - May 2025, San Francisco, CA):
  - Architected “Books Done Right,” an AI bookkeeping application using Next.js, React, Node.js, Python, LLM tool calling, and the QuickBooks API.
  - Implemented an LLM-based natural-language interface that reduced manual bookkeeping effort by 80% while achieving 95%+ classification accuracy.
  - Translated bookkeeping workflows and edge cases into model, prompt, API, and UI behavior for non-technical users.
- **Full-Stack Software Engineer at CmdTower** (Apr 2024 - Aug 2024, San Francisco, CA):
  - Built a vertical SaaS platform with TypeScript-based onboarding and authentication, a Django backend on AWS Elastic Beanstalk, and AWS CodePipeline CI/CD for deployment.

## Research

- **Explainable AI for Brain Tumor Diagnosis** (MESA Research Program, Sonoma State University): Built and evaluated CNN image-classification models (VGG16, VGG19, ResNet-50, MobileNet) on the Br-H35 brain-tumor dataset; applied LIME, SHAP, and Grad-CAM for interpretability and evaluated performance with AUC-ROC curves and confusion matrices.

## Education

- **Sonoma State University**: B.S. Computer Science (Aug 2023 - May 2025). Graduated with Honors; International Student Scholarship recipient (May 2023). Relevant coursework: Database Systems, Data Structures & Algorithms, Computer Architecture, Software Design & Development, Computer Networking, Analysis of Algorithms.
- **City College of San Francisco**: A.S. Computer Science (Aug 2021 - May 2023).

## Skills

- **AI / ML**: LLM integration, prompt engineering, tool calling, RAG, TensorFlow, PyTorch, LIME, SHAP, Grad-CAM
- **Frontend**: React, Next.js, TypeScript, Tailwind CSS, HTML/CSS
- **Backend**: Node.js, Python, Java, Django, Express.js, REST APIs, QuickBooks API
- **Cloud & Infra**: AWS (Certified Developer), Docker, CI/CD, MongoDB, PostgreSQL, Vercel
- **Ways of working**: Customer discovery, technical scoping under ambiguity, technical diligence, executive communication, ROI framing, Git, Agile

## Blog

Technical articles on AI, React, Next.js, TypeScript, and software development.
All posts are available in LLM-friendly markdown: append `.md` to any blog post
URL, or send `Accept: text/markdown`.

### Articles

**AI:**
- [Nex-N2.5-mini Review: A Compact Agentic Model for Computer Use and Web Tasks](https://samkhan.net/blog/nex-n2-5-mini-review-agentic-model-2026): Evaluating Nex-AGI's smallest agentic model for lightweight workflows, computer use, and web browsing tasks - free, open-source, and capable for experimentation.
- [MiniCPM5-2B Review: 2B Parameter Model Excels in Coding and Agent Tasks](https://samkhan.net/blog/minicpm5-2b-review-2026): Review of OpenBMB's MiniCPM5-2B, a dense 2B Transformer model achieving 2B-class SOTA performance with strong capabilities in coding, mathematics, and tool use.
- [Spark-X2.5-4B Review: Hybrid Architecture for Agentic Workflows](https://samkhan.net/blog/spark-x2-5-4b-review-hybrid-architecture-agentic-workflows): Analysis of Spark-X2.5-4B's hybrid attention architecture, benchmark performance in agentic tasks, and practical deployment considerations for developers.
- [GLM-5.3 Review: Z.ai's New Open Weights Model for Coding and Agentic Workflows](https://samkhan.net/blog/glm-5-3-review-2026): Comprehensive review of GLM-5.3, Z.ai's latest open weights Mixture-of-Experts model showing strong performance on coding benchmarks and cybersecurity tasks.
- [Qwen3.8-Flash-Next Review: Alibaba's Latest Multimodal Model for Vision and Text Tasks](https://samkhan.net/blog/qwen3-8-flash-next-review-2026): Review of Qwen3.8-Flash-Next, Alibaba's experimental 125B parameter model with 6B activated parameters, featuring hybrid attention, n-gram embeddings, and extensive benchmark results.
- [GLM-5.3-Flash vs Qwen3.8-Flash-Next: Comparing Leading Efficient LLMs for Agentic Workflows](https://samkhan.net/blog/glm-5-3-flash-vs-qwen3-8-flash-next-comparison-2026): Head-to-head comparison of Z.ai's GLM-5.3-Flash and Qwen's Qwen3.8-Flash-Next, two Mixture-of-Experts models optimized for cost-effective agentic AI and long-context processing.
- [DeepSeek-V4-Flash-Vision-Exp Review: A Multimodal Frontier Model](https://samkhan.net/blog/deepseek-v4-flash-vision-exp-review-2026): First look at DeepSeek's experimental multimodal model showing strong text agent capabilities and improved vision understanding, benchmarked against Opus 4.8.
- [Tiel Coder 35B Review: Local Coding Model Matches Opus 4.6 Medium on SWE-Bench Live](https://samkhan.net/blog/tiel-coder-35b-review-specialized-coding-model-2026): A dynamically quantized MoE model that matches Opus 4.6 medium on coding benchmarks while excelling in multi-turn conversations for agentic workflows
- [OpenAI Python SDK v3.6.0 Adds Compute Units to Usage Tracking](https://samkhan.net/blog/openai-python-sdk-v3-6-0-compute-units): The latest OpenAI Python SDK release v3.6.0 introduces compute_units tracking in API response usage, enabling developers to monitor computational resource consumption alongside token counts.
- [GPT-5.2 Technical Review: OpenAI's Best Model](https://samkhan.net/blog/gpt-5-2-technical-review): GPT-5.2 lands in the middle of an unprecedented AI arms race. With a 400K context window, 30% fewer hallucinations, and the first 90%+ ARC-AGI score, here's what developers need to know about OpenAI's most capable model yet.
- [GPT-5.2 vs Gemini 3: Real Coding Test Results](https://samkhan.net/blog/gpt-5-2-vs-gemini-3-tested): GPT-5.2 dropped yesterday. I put it head-to-head against Gemini 3 on real development tasks: debugging, refactoring, API design, and more. The results surprised me.
- [Grok 4.1 Review: xAI's Latest Model Tested](https://samkhan.net/blog/grok-4-1-review-2025): Complete review of Grok 4.1, xAI's latest language model. Benchmark performance, real-world testing, cost analysis, and comparison with GPT-5.2, Claude Opus 4.5, and other frontier models.
- [Mistral Devstral 2 Review: Agentic Coding Model](https://samkhan.net/blog/mistral-devstral-2-agentic-coding-review-2025): Mistral Devstral 2 is purpose-built for agentic coding workflows. Complete review of its coding capabilities, agent performance, and how it compares to GPT-5.2, Claude Opus 4.5, and specialized coding tools.
- [Mistral Large 3 Review: GPT-5.2 Competitor?](https://samkhan.net/blog/mistral-large-3-review-2025): Complete review of Mistral Large 3 (2512), the latest frontier model from Mistral AI. Benchmark comparisons, coding performance, cost analysis, and real-world testing against GPT-5.2 and Claude Opus 4.5.
- [Prompt Engineering Guide: Better AI Outputs](https://samkhan.net/blog/the-art-of-prompt-engineering): LLMs are prediction engines. The quality of your output depends entirely on how you ask. From basic techniques to advanced strategies like Chain-of-Thought and ReAct, plus production-level patterns from Parahelp's XML-structured prompts.
- [OLMo 3 32B Think Review: Best Open-Source LLM](https://samkhan.net/blog/allenai-olmo-3-32b-think-review-2025): Complete review of OLMo 3 32B Think, AllenAI's open-source reasoning model. Benchmark performance, self-hosting guide, cost analysis, and comparison with GPT-5.2, DeepSeek-V3.2-Speciale, and other open-source models.
- [Amazon Nova 2 Lite Review: AWS AI Model Guide](https://samkhan.net/blog/amazon-nova-2-lite-review-2025): Complete technical review of Amazon Nova 2 Lite, AWS's latest language model. Benchmark performance, AWS integration, cost analysis, and comparison with GPT-5.2, Claude, and other frontier models.
- [Claude Opus 4.5: Complete Developer Review](https://samkhan.net/blog/claude-opus-4-5-complete-review-2025): Comprehensive review of Claude Opus 4.5, Anthropic's latest frontier model. Complete capabilities analysis, benchmark performance, real-world testing, cost evaluation, and developer use case recommendations.
- [DeepSeek V3.2 Speciale Review: Free GPT-5 Rival](https://samkhan.net/blog/deepseek-v3-2-speciale-review-2025): DeepSeek-V3.2-Speciale delivers frontier-level reasoning capabilities with open weights. Complete technical review, benchmark comparisons, and developer testing of this free alternative to GPT-5.2 and Claude Opus 4.5.
- [GPT-5.1 Codex Max vs Claude Opus 4.5 for Coding](https://samkhan.net/blog/gpt-5-1-codex-max-coding-review-2025): Complete comparison of GPT-5.1-Codex-Max and Claude Opus 4.5 for coding tasks. Benchmark performance, real-world coding tests, cost analysis, and developer workflow recommendations.
- [GPT-5.2 Developer Review: First Look (Dec 2025)](https://samkhan.net/blog/gpt-5-2-developer-review): A comprehensive developer review of GPT-5.2 released December 11, 2025. Comparing Instant, Thinking, and Pro variants with benchmarks, pricing, and real-world coding performance.

**Developer Tools:**
- [Running gh-aw agents on NVIDIA free NIM endpoints](https://samkhan.net/blog/running-gh-aw-agents-on-nvidia-free-nim-endpoints): Wiring GitHub Agentic Workflows to use NVIDIA's free hosted NIM endpoints via Copilot BYOK engine mode
- [Playwright MCP Setup Guide for Claude Code](https://samkhan.net/blog/playwright-mcp-claude-code-browser-automation): How to set up Playwright MCP with Claude Code for AI-powered browser automation. Control Chrome, fill forms, scrape data, and automate web testing with natural language commands.
- [How I Cut My Debugging Time in Half as a React Developer](https://samkhan.net/blog/how-i-cut-debugging-time-in-half): After tracking my debugging sessions for a month, I discovered that 80% of my bugs fell into 5 categories. Here's what I changed and the tools that actually made a difference.
- [Cursor vs Copilot vs Claude Code: 2025 Comparison](https://samkhan.net/blog/ai-coding-tools-comparison-2025): The definitive comparison of AI coding assistants in 2025. Learn which tool is best for your workflow: Cursor, GitHub Copilot, or Claude Code, with pricing, features, and real-world recommendations.

**Web Development:**
- [TypeScript Type vs Interface: When to Use Each (2025)](https://samkhan.net/blog/typescript-type-vs-interface-2025): The definitive guide to choosing between TypeScript's type and interface. Learn when to use each with real-world examples, common patterns, and practical decision guidelines.
- [useEffect Guide: Fix Common React Problems 2025](https://samkhan.net/blog/useeffect-complete-guide-2025): Master React's useEffect hook with practical solutions to infinite loops, cleanup issues, dependency arrays, and more. Real-world examples and debugging strategies.

**Projects:**
- [How I Built This Portfolio with Next.js 16 and Tailwind CSS](https://samkhan.net/blog/building-this-portfolio): A deep dive into the architecture, design decisions, and optimizations that went into building my personal portfolio site. From SEO to accessibility, learn what goes into a modern developer portfolio.

## Contact

- Email: khansam@sonoma.edu
- Website: https://samkhan.net
- LinkedIn: https://linkedin.com/in/sameerkhansf
- GitHub: https://github.com/sameerkhansf
- X: https://x.com/sameerkhan_sf