Akshat Sahu
Building production-grade AI agent systems. Specializing in multi-agent orchestration, RAG pipelines, and enterprise AI.
const engineer = {
name: "Akshat",
focus: "AI Agents",
stack: [
"Multi-Agentic Workflows",
"DeepAgents",
"LangGraph",
"RAG",
"MCP"
],
degree: "MS CS",
gpa: 3.9
}
Get to know me
About Me
Software Engineer
Founding Engineer @ Future Path AI
I'm a Software Engineer with a Master's in Computer Science, passionate about building and startups. I specialize in production-grade AI agent systems at scale — multi-agent orchestration on LangGraph and DeepAgents, hybrid retrieval RAG (pgvector + BM25 via RRF) with cross-encoder reranking, and MCP-driven integrations spanning 80+ enterprise tools. At Future Path AI, I've architected autonomous agents for Fortune 500 pharma and finance clients, with rigorous DeepEval / RAGAS / Langfuse evaluation pipelines guiding model selection across OpenAI, Anthropic, and Gemini.
Professional journey
Work Experience
Founding Software Engineer
Future Path AI
- Architected AI agent infrastructure for an enterprise platform autonomously resolving business and IT operations at Fortune 500 pharma and finance clients — built the multimodal RAG layer with hybrid retrieval (pgvector + BM25 via RRF), cross-encoder reranking, and a vendor-agnostic embedding layer over 10K+ documents.
- Built agent and RAG evaluation pipelines using DeepEval, RAGAS, and Langfuse — 200+ per-client test cases across tool-use reliability, task completion, faithfulness, and context precision — driving model selection and prompt iteration across OpenAI, Anthropic, and Gemini.
- Developed a multi-agent orchestration platform on LangGraph and DeepAgents with manifest-driven agent definitions (config-as-code) — leveraging planning loops, sub-agent delegation, and file-system context management to autonomously execute multi-step IT and business operations workflows.
- Architected an MCP orchestration layer spanning 80+ enterprise integrations (ServiceNow, Kubernetes, Splunk, Dynatrace, Azure, SAP, Salesforce) with multi-tenant credential injection and dynamic per-query skill selection — powering LangGraph autonomous AI and chat agents handling 100+ tickets daily across ITSM, observability, and ERP workflows.
Software Engineering Intern
Nagarro
- Built a full-stack product community web application end-to-end using Java, Spring Boot, Hibernate, MySQL, and Angular, delivering modular services for user registration, product browsing, and review management.
- Designed and implemented RESTful APIs with Spring Boot for user authentication, product search, and review workflows, applying caching and connection pooling to optimize response latency.
- Engineered Hibernate ORM data models with MySQL — relational schema design, lazy loading, and indexed queries for product and review entities.
- Developed responsive Angular interfaces with component-based architecture, reactive forms, and client-side state management for product browsing and review submission flows.
Academic background
Education
Stevens Institute of Technology
MS · GraduateMaster of Science in Computer Science
Manipal University Jaipur
BE · UndergraduateBachelor of Engineering in Information Technology
Technologies I work with
Skills & Expertise
Languages
7 skills
Recent work
Projects
Analyzes whether your AI coding agent is actually shipping code by correlating Claude Code token usage with git commits. Reads local session files, matches with git history by timestamp, and generates an ROI dashboard — zero config.
Cost per Commit: calculates token expenses per AI-assisted commit to surface the true dollar cost of each code change.
Line Survival Rate: tracks the percentage of AI-written code still present after 24 hours, distinguishing genuine contributions from throwaway drafts.
Orphaned Sessions: detects unproductive Claude Code sessions with no matching git activity, highlighting wasted spend.
Solo AI coaching platform helping startup founders refine pitch decks through automated analysis, real-time voice and chat investor simulation, and feedback grounded in Y Combinator pitch heuristics.
Built PitchPilot, a solo AI coaching platform helping startup founders refine pitch decks through automated analysis, real-time voice and chat investor simulation via the OpenAI Realtime API, and feedback grounded in Y Combinator pitch heuristics.
Designed multi-agent workflows on FastAPI and LangGraph with persistent state, ingesting PDF/PPTX/DOCX decks via a document parsing pipeline backed by PostgreSQL, Prisma, and Supabase Storage.
Engineered an LLM-driven PDF and PPTX generation pipeline using few-shot prompting against curated benchmark decks to produce polished revisions and coaching reports for founders.
Full-stack pet services platform supporting pet adoption, provider matching, and appointment scheduling, with secure authentication and community-driven review features.
Built a full-stack pet services platform on Node.js, Express, MongoDB, and React — supporting pet adoption, provider matching, and appointment scheduling through Mongoose-modeled schemas and RESTful APIs.
Implemented secure authentication with bcrypt salted hashing, express-session server stores, and OTP-based password recovery with token expiration and rate-limited endpoints.
Built community features — threaded comments, review analytics, and booking flows — powered by MongoDB aggregation pipelines for review scoring and ElasticEmail for transactional notifications.
Contact
Get In Touch
Let's build something together.
Have a project in mind or want to discuss opportunities? I'd love to hear from you.
“I'm particularly interested in early-stage AI infrastructure, multi-agent systems, and founding engineering roles.”