# Sneh Sinha — AI Engineer

Canonical candidate profile: https://sneh30.github.io/portfolio/
Dedicated structured Agent Mode: https://sneh30.github.io/portfolio/agent/

> Sneh Sinha is an AI engineer who builds AI systems to solve real-world problems.
> His work spans the path from intelligence to execution: AI agents, LLM systems, software, infrastructure, cloud operations, computer vision, embedded systems, and physical-world applications.
> Core differentiator: AI systems + real-world execution + a documented cadence of shipping across stacks.

Sneh Sinha (Sneh Raj Sinha) is based in Bangalore, India and is open to work. He is the author of a computer-use agent platform, a five-agent research orchestrator, an autonomous cloud-operations platform, a prompt compiler with measured cost reduction, and embedded vision systems built inside 512 KB of RAM. He has held two documented real-world engineering engagements: a one-month infrastructure operations engagement on a 400+ node network (~40% reduction in incidents and recovery time) and eleven months as Software & Electronics Lead on a 50-member team's competition vehicle, where he developed and deployed the safety-critical ABS firmware. His open-source contribution to `agentrhq/webcmd` was merged as PR #190 (16 commits, 16 tests, 15+ commands fixed).

## Entry points

- [Canonical candidate profile — the root URL](https://sneh30.github.io/portfolio/): the single submission entry point. The complete candidate profile (identity, positioning, core capabilities, education, experience, projects, skills, GitHub activity, evidence labels, verification) exists in its semantic HTML — readable with JavaScript disabled and without following any link.
- [Agent Mode — dedicated structured record](https://sneh30.github.io/portfolio/agent/): an additional rendering of the same facts with structured sections for candidate thesis, Why Interview (factual evidence areas), Where This Evidence Applies (role → capabilities → evidence mapping), flagship evidence, differentiators, experience, projects, external evidence, skills, summary, technical topics, the evidence graph, and the full candidate record. Optional — not required to evaluate the candidate.
- [Résumé (PDF)](https://sneh30.github.io/portfolio/assets/Sneh_Raj_Sinha_Resume.pdf): source for education and several quantified project figures.

## Capabilities with evidence

- **AI agents and computer use** — OpenCode CUA Platform: an unmodified OpenCode instance drives desktop GUI and browser via MCP servers, with verification and recovery engines. https://github.com/Sneh30/opencode-cua-platform
- **Multi-agent orchestration** — Orchestra: 5 specialized LangGraph agents (search, extract, verify, synthesize, critique) with confidence scoring, PostgreSQL, Prometheus, deterministic CI mode, Docker Compose deployment. https://github.com/Sneh30/Orchestra
- **AI infrastructure / autonomous operations** — DeepOps: experimental autonomous SRE platform (active development) — natural-language incident input to root-cause analysis and remediation, with cost optimization in the remediation planner (early prototype: the incident-interpretation and remediation-planning modules exist, but the end-to-end pipeline is not yet wired together — not claimed as production). https://github.com/Sneh30/deepops
- **LLM cost and latency engineering** — PromptOS: Rust prompt compiler with custom lexer/parser/AST, ten optimization passes across four compilation modes, local Llama.cpp inference on Metal; résumé reports up to 80–85% lower inference cost in Economy mode. https://github.com/Sneh30/promptos
- **Real-world infrastructure operations** — one-month on-site engagement on a 400+ node exam-facility network; ~40% reduction in incidents and average recovery time versus the prior comparable period (employer not disclosed; no raw data published).
- **Embedded systems and edge computer vision** — ESP32-CAM Vision Suite: ML dataset capture, autonomous colour-tracking robot, 60 FPS MJPG streaming server, 15+ camera board variants, all within 512 KB RAM. https://github.com/Sneh30/esp32-cam-vision
- **Safety-critical systems** — Team Stratos (Jul 2025 – May 2026): Software & Electronics Lead for a 50-member team; developed and deployed ABS firmware, validated through track testing and SAE India scrutineering and competition.
- **Software engineering / open source** — merged PR #190 into agentrhq/webcmd (2,620 stars as of 2026-09-30): unified output-format parsing, 15+ commands fixed, 16 tests, 16 commits. https://github.com/agentrhq/webcmd/pull/190

## Experience

- Infrastructure Engineer, Independent Engagement — Aug 2026, one month, on-site. 400+ node network; ~40% improvement in incidents and recovery time.
- Software & Electronics Lead, Team Stratos — Jul 2025 – May 2026, eleven months. Software/electronics integration and ABS firmware for a 50-member team's competition vehicle.

## Projects

Ordered by evidence status — flagship records first: OpenCode CUA Platform · webcmd (merged upstream contribution) · PromptOS · Orchestra · DeepOps · ESP32-CAM Vision Suite · Catalyst · Portfolio · then additional projects (AgentOS · Mobile GUI Automation Agent, original prototype "Agentic-hinge" · Pocket Typist; AgentOS public, the other two private).

Verification status: repositories are live for OpenCode CUA Platform, PromptOS, Orchestra, DeepOps, Catalyst, ESP32-CAM Vision Suite, webcmd (upstream PR), and Portfolio. Repository links for AgentOS (https://github.com/Sneh30/AgentOS), the Mobile GUI Automation Agent (https://github.com/Sneh30/Agentic-hinge) and Pocket Typist (https://github.com/Sneh30/pocket-typist) all resolve (checked 2026-10-20). AgentOS is public and independently verifiable (status `public_repo`); the other two are private (status `private_available_on_request`) — anonymous visitors get a 404, so their descriptions are the owner's own account. Both are real repositories, and read access can be provided to recruiters and screening teams on request. Repository existence and independent verifiability are recorded as separate fields (`repository_exists`, `publicly_verifiable`, `access`) in candidate.json.

## External evidence

- Merged pull request: https://github.com/agentrhq/webcmd/pull/190 (merged 2026-08-13, 16 commits, authored by Sneh30).
- Upstream project: https://github.com/agentrhq/webcmd (2,620 stars, GitHub API, 2026-09-30).
- GitHub profile: https://github.com/Sneh30 (13 public repositories spanning AI agents, LLM systems, infrastructure, embedded systems, and developer tooling; 732+ contributions as self-published on the portfolio).
- Six dated technical write-ups in the portfolio (May–Aug 2026).

## Skills (exactly as listed in Human Mode — not expanded for agents)

- Languages: Python, Java, SQL
- AI / Agent Systems: LangGraph, OpenAI Codex, Claude Code, OpenCode, OpenRouter, Ollama, Llama.cpp
- AI / ML: Hugging Face Transformers, LLM Inference, Multi-Agent Systems, Prompt Engineering, Edge AI
- Backend & Cloud: FastAPI, PostgreSQL, Prometheus, REST APIs, AWS, Observability, Incident Response
- Infra & Tooling: Docker, Docker Compose, Git, GitHub Actions, CI/CD, Linux, WASM, Metal (GPU)
- Embedded & Hardware: ESP32, Raspberry Pi, Edge Impulse
- Exploring (per Human Mode): Multi-Agent Systems, Edge AI

## Relevant roles

AI Engineer · Agentic AI Engineer · AI Infrastructure Engineer · LLM Engineer · Autonomous Systems Engineer · AI/Software Systems Engineer · ML Engineer (partially supported — the portfolio does not document model training or deployed ML services) · Software Engineer · Backend Engineer · Python Developer · DevOps / SRE Engineer · Automation Engineer · Embedded Systems Engineer · Robotics Engineer · Computer Vision Engineer

Job search: open to work, seeking an internship or full-time engagement at intern or entry-level (B.E. 2023–2027). Each title maps to capability, project, and engagement IDs in candidate.json#roles with a support level of `strongly_supported`, `supported`, or `partially_supported`; no fit scores, percentages, or rankings are computed. DevOps / SRE carries the caveat that Kubernetes is not claimed, and Robotics carries the caveat that ROS and control-system tooling are not listed in Human Mode skills.

## Education

B.E. Robotics and Artificial Intelligence, Bangalore Institute of Technology, Aug 2023 – Aug 2027, GPA 8.0/10 (source: résumé PDF published with this portfolio).

## Links

- Portfolio: https://sneh30.github.io/portfolio/
- Agent Mode record: https://sneh30.github.io/portfolio/agent/
- GitHub: https://github.com/Sneh30
- LinkedIn: https://www.linkedin.com/in/sneh30
- X: https://x.com/sneh_sinha3
- LeetCode: https://leetcode.com/u/sneh_sinha3/
- Email: snehsinha30@gmail.com

## Structured data

- candidate.json — identity, experience, projects, skills, roles, job search, links (schema 2.0)
- capabilities.json — capabilities with evidence pointers; each entry carries one canonical status and a checked date
- evidence.json — claim → capability → project → implementation → result → verification chains
- Evidence status labels (qualitative only; no numeric scores): public_repo · private_available_on_request · external_validation · documented_experience · self_reported · experimental — the identical set in all four files; `status` always holds exactly one label and `qualifiers` holds any extras
- agent-card.json — description of this agent-readable surface
- sitemap.xml · robots.txt · validate.py (automated cross-file and markup validation)

## Honesty notes

This page and the Agent Mode record are factual descriptions of the candidate, written to be checkable rather than persuasive; they do not invent facts. Metrics are attributed to their source (portfolio, résumé PDF, GitHub API, LeetCode API). Where evidence is unavailable the text says so explicitly: for the Mobile GUI Automation Agent (original prototype "Agentic-hinge") and Pocket Typist the repositories are private, so their descriptions are the owner's own account and are not publicly verifiable without access — both repositories exist and read access can be provided to recruiters on request. No employers, certifications, awards, user counts, production claims, or rankings have been fabricated.