Skip to content
View akira231097's full-sized avatar
🎯
Focusing
🎯
Focusing
1 follower · 0 following

Block or report akira231097

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
akira231097/README.md

Hi, I'm Sarath Chandra Gentela

AI Agent Harness & LLM Systems Engineer

I build production AI agents end-to-end — controlled LLM orchestration runtimes, retrieval systems, media pipelines, and inference-aware deployments.

LinkedIn Email Phone Location Profile views

About

I'm an AI Engineer focused on agentic systems, retrieval, and LLM inference. At Spice (thespice.ai) I built the company's AI products and agents from the ground up — including Pulse / Lucid, a creator "content chief-of-staff" agent harness (try-pulse.ai), and AskSpice, conversational search over 10,000+ podcast episodes and 1M+ searchable chunks.

My strongest work spans agent-harness engineering, LLM orchestration, tool execution, Retrieval-Augmented Generation, hybrid search, durable memory, automated decision loops, and LLM inference engineering.

🎓 M.S. Data Analytics — University of Illinois Springfield (4.0 GPA) 📍 Austin, TX · 💬 open to AI / LLM / agent engineering roles

Featured Projects

Project What it does Stack
EchoFind Memory-aware conversational RAG engine that returns the single best podcast clip or episode for a natural-language question FastAPI · Gemini · Pinecone · Cohere · Postgres
Clip'O'pedia Mention-driven hybrid-RAG clip recommender on a LangGraph pipeline — runs fully offline with zero API keys LangGraph · HyDE · RRF · hexagonal
ReelForge AI post-production pipeline: turns a talking-head clip + transcript into a polished 9:16 captioned reel Gemini · FFmpeg · MediaPipe · MoviePy
Artha Council Staged, multi-model AI investment committee with a fail-closed OpenClaw agentic broker bridge GPT-5.5 · Gemini · Claude · MCP
Commitment Decay Engine Turns meeting transcripts into a commitment ledger and reconciles follow-through Python · CLI · LLM-ready

New here? Start with EchoFind (memory-aware RAG) or Artha Council (multi-model agents + fail-closed agentic execution).

Tech Stack

Languages

Python TypeScript Node.js SQL

AI · Agents · Retrieval

OpenAI Gemini Claude LangGraph Pinecone Cohere RAG

Backend · Infrastructure

FastAPI PostgreSQL AWS Docker Terraform GitHub Actions

LLM Inference

vLLM SGLang TensorRT--LLM Quantization

Pinned Loading

  1. artha-council artha-council Public

    Staged, multi-model AI investment-committee engine. Scans a 1000+ US-stock universe through a promotion funnel, has GPT, Gemini and Claude analysts debate finalists, then applies deterministic, fai…

    Python 1

  2. clipopedia clipopedia Public

    Mention-driven hybrid-RAG assistant that recommends the single best podcast clip for a question. Query understanding + HyDE + hybrid dense/sparse search + reciprocal rank fusion + cross-encoder rer…

    Python

  3. commitment-decay-engine commitment-decay-engine Public

    Supportive, CLI-first commitment tracking engine. Extracts action items from meeting transcripts and chat into a human-readable markdown ledger, reconciles them against evidence (tickets, PRs, upda…

    Python

  4. echofind echofind Public

    Memory-aware conversational RAG engine that returns the single best podcast clip or episode for a natural-language question. LLM intent router, HyDE + hybrid dense/sparse retrieval, multi-level RRF…

    Python

  5. reelforge reelforge Public

    AI post-production pipeline that turns a talking-head clip plus its word-level transcript into a polished 9:16 reel: Gemini-planned edit, Sora/Pexels B-roll with quality-gated routing, MediaPipe fa…

    Python