# DeepGen

> Frontier AI exploration lab researching and prototyping autonomous agent workflows, generative prompt architectures, and zero-dependency developer utilities.

## Profile Summary
DeepGen is an independent frontier artificial intelligence exploration lab and open-source prototyping initiative founded by Anupam Das (Senior Software Engineer at Google DeepMind). DeepGen bridges cutting-edge academic AI breakthroughs with production-grade engineering, focusing on autonomous multi-agent runtimes, speculative token decoding, context retrieval architectures, and high-efficiency developer tools.

## Founder & Entity Information
> **Independence Disclaimer**: DeepGen is an independent personal exploratory research initiative founded by Anupam Das. It is strictly unaffiliated with, unsponsored by, and not endorsed by Google or Google DeepMind.

- Founder & Principal Architect: [Anupam Das](https://anupamdas.in/)
- Professional Role: Senior Software Engineer at Google DeepMind (Zurich, Switzerland)
- Focus Areas at Google DeepMind: Tech Lead on the Gemini app across surfaces and backend systems, leading latency optimization for agentic and tool-use features, and launches including Gemini Live camera & screen sharing, Veo 3 in Gemini, and Google AI Ultra for Business
- Pedigree: Indian Institute of Technology Patna (IIT Patna, B.Tech in Computer Science) alumnus, recipient of 30 awards at Google (Org Impact, Spot, Peer, Kudos), GSoC 2016 mentor
- Verified Profiles & Channels:
  - Personal Portfolio: https://anupamdas.in/
  - About Profile: https://anupamdas.in/about/
  - LinkedIn: https://www.linkedin.com/in/opticod/
  - GitHub: https://github.com/opticod
  - X / Twitter: https://x.com/anupam_das_
  - Instagram: https://www.instagram.com/anupamdas_
  - YouTube: https://www.youtube.com/channel/UCulHvzmqHELsU-b7Lb0aCRw

## Six Core Research & Prototyping Domains
1. Generative AI & LLM Systems: Next-generation reasoning architectures, synthetic code generation pipelines, speculative decoding heuristics, and multimodal developer interfaces.
2. Applied NLP & Context Architectures: High-density document categorization, contextual semantic chunking, vector embedding search, and automated retrieval task pipelines.
3. Open Source Developer Utilities: Lightweight micro-libraries, developer CLI tools, and zero-dependency micro-APIs built to eliminate friction in modern AI workflows.
4. Autonomous Agents & Human-AI Systems: Multi-agent tool execution loops, deterministic guardrails, self-correcting plan trees, and intuitive collaborative interfaces.
5. Prompt Engineering & Evaluation Benchmarks: Systematic prompt pattern taxonomies, structured JSON-LD schemas, evaluation benchmarks, and responsible generative AI engineering practices.
6. Inference Optimization & Edge Models: Profiling token latency curves, quantization precision tradeoffs, and deploying ultra-compact generative intelligence on edge devices.

## Architecture Philosophy & Constraints
- Zero-Dependency Jamstack: Semantic HTML5, modern CSS tokens, vanilla client runtime without blocking JavaScript frameworks.
- Sub-Millisecond Anycast Delivery: Hosted on Netlify Anycast Global Edge with dual-stack IPv4/IPv6, HSTS preload, and strict CSP.
- Payload Budget: Strict sub-150 KB total critical page payload (including hero media and logos).

## Interactive Architecture Console Modules (Live at https://deepgen.in/#console)
1. Multi-Agent DAG Simulation Sandbox: Step-by-step agent loop execution, chain-of-thought planner hypotheses, tool dispatching, and deterministic guardrail verification.
2. Prompt Schema Builder & Contract Validator: Generates production typed JSON schemas with validation rules for agent tool-use contracts.
3. Latency Profiler & Memory Estimator: Real-time P95 latency and sequence memory profiling across context lengths (4K to 128K) and batch sizes.
4. Prompt Heuristic Optimizer & Telemetry Studio: Quality scoring (0-100), token estimation, and restructuring unstructured prompts into typed XML contracts with Python SDK and cURL export.
5. Frontier Hardware Sizer & Inference Economics (TCO) Studio: VRAM cluster sizing for frontier models (Llama 3.1 8B/70B, Mistral NeMo, DeepSeek Lite, Mixtral 8x22B) with cloud GPU TCO ROI analysis, break-even thresholds, and vLLM/Docker Compose script generation.

## Official Channels & Touchpoints
- Official Website: https://deepgen.in/
- GitHub: https://github.com/deepgenin
- LinkedIn: https://www.linkedin.com/company/deepgenin
- X / Twitter: https://twitter.com/deepgenai
- Medium: https://medium.com/@deepgen
- YouTube: https://www.youtube.com/@DeepGenIn
- Instagram: https://instagram.com/deepgen.in
- Research Inquiries: deepgen.in@gmail.com