I am computer science grad, I have 2 years of experience in backend development. Worked in an early stage startup

github

Poras Shrivas  • Peerlist

Built an AI counsellor agent that runs entirely inside Instagram DMs. Not a chatbot wrapper around an LLM, but a 20-node LangGraph state machine handling intent classification, emotional state routing, real-time personalization, payment gating, and multimodal image analysis, all while feeling like a real conversation. -Dual-LLM architecture. Groq (Llama 3.3 70B) handles sub-second intent classification, Gemini handles creative response generation, and there's a triple-layer fallback so the agent never goes silent. -Deterministic computation layer. Every response is grounded in real calculated data, not LLM guesswork. It's computed fresh per conversation and injected into the prompt as structured context. - Message debouncing. Users often send bursts of messages within seconds of each other. The webhook pushes each one to a Redis inbox with a setnx lock, and the worker drains the full queue before running the graph. That means one coherent reply instead of the agent responding to four fragments separately. - Human-like delivery. Responses are chunked and delivered as separate DM bubbles with WPM-based typing delays through a job queue, and Redis send-locks stop messages from arriving out of order. - Semantic memory. An LLM updates a structured user-facts object after every turn (stated problem, emotional triggers, prior insights given), with anti-repetition guards so the agent doesn't repeat itself across long conversations. - Emotional state handling. A dedicated classifier detects sustained distress and switches the agent into a grounding mode: predictions get suppressed, tone shifts toward warmth, and a professional referral gets surfaced once per session. - Triple-idempotent payments. Razorpay webhooks are verified with HMAC and deduplicated at the Redis lock, session state, and DB level. A successful payment triggers a proactive follow-up through the same agent pipeline.

AI Counsellor Agent | Python · FastAPI · Lang

Built an end-to-end autonomous pipeline that generates and publishes short-form video content daily, with zero manual intervention from data ingestion to social publishing. The system pulls real-time trend signals (Reddit API, Google Trends RSS), scores them for engagement potential, and feeds the highest-scoring angle into an LLM pipeline (Groq/LLaMA-3) constrained to output structured JSON: hooks, paced scripts, visual cues, and captions. That JSON drives a React + Remotion rendering pipeline that programmatically builds and exports a 9:16 video, which is then uploaded to Cloudinary and auto-published via the Meta Graph API. Key engineering decisions: Structured-output prompting to make LLM responses reliably parseable by downstream code, not just human-readable Decoupled architecture, so data scoring, generation, rendering, and publishing run as independent stages that can fail or retry without breaking the pipeline Fully code-driven video rendering (React/TypeScript via Remotion) instead of manual editing, enabling daily scale Automated via GitHub Actions for unattended daily runs Stack: Python, React, TypeScript, Remotion, Groq API (LLaMA-3), Reddit API, Google Trends, Meta Graph API, Cloudinary, GitHub Actions

Autonomous Content Automation Engine

Built a browser-based file transfer tool after discovering the practical limits of pure P2P architectures and redesigning it around a faster, more reliable approach. * Started with WebRTC data channels for direct browser-to-browser transfers. Benchmarked against major P2P file-sharing tools and found consistent throughput limits caused by browser networking constraints rather than implementation quality. * Identified additional reliability issues with NAT traversal, carrier-grade NAT (CGNAT), and asymmetric consumer internet connections, which frequently degrade real-world P2P performance. * Redesigned the transfer path using Cloudflare Durable Objects to maintain stateful edge sessions and stream files through the nearest edge location, significantly improving transfer speed and reliability while preserving a browser-only experience. * Kept pure P2P as an optional mode for privacy-focused users who prefer direct peer-to-peer transfers. * Implemented resumable file transfers by tracking byte offsets, allowing interrupted transfers to continue from the last completed position instead of restarting. * Grew the platform to 500+ organic users across multiple countries with zero marketing spend.