Career case file / Saksham Pathak

Creator.
Coder.Engineer.

From a 100K+ devotional content creator to a Generative AI Engineer at Zencia AI — the story of how Saksham Pathak, known online as Parthmax, rebuilt himself around voice agents, agentic AI, and production LLM systems.

Saksham Pathak (Parthmax) Creator → AI Engineer
100K+Digital following
M.Sc. AI/MLIIIT Lucknow
Zencia AICurrent role
350+LeetCode DSA solved
Career case file / Chapters

Five chapters, one thread.

Every chapter looks unrelated to the last — devotional editing, pure math, freelance AI evaluation, shipped open-source tools, production voice AI — until you see the thread running through all of it: understanding what makes something actually work for people.

CASE / 01 2021 – 2024

Creative Genesis

While studying B.Sc. Mathematics at Dr. Ram Manohar Lohia Avadh University, Saksham built a 100K+ following on Instagram and YouTube editing devotional content — an early lesson in what makes content resonate at scale.

CASE / 02 Aug 2024

The Pivot

Began an M.Sc. in Artificial Intelligence & Machine Learning at IIIT Lucknow, trading the editing timeline for transformer architectures and systems thinking.

CASE / 03 Oct 2024 – Mar 2025

Outlier

Freelance AI developer work in LLM evaluation, prompt engineering, and response-quality assessment — judging how well models actually reason, from the inside.

CASE / 04 2025

Shipping in Public

FALCON, DocuMind AI, and parth-dl ship under the Parthmax name — public on GitHub and Hugging Face, not just claimed but checkable.

CASE / 05 Aug 2025 – Present

Zencia AI

Generative AI Engineer, building real-time conversational voice agents and agentic AI systems in production — M.Sc. completed Jun 2026 alongside the role.

Technical arsenal / Full stack

Everything the roles above are built on.

Not a buzzword list — this is the actual stack behind FALCON, DocuMind AI, CodePilot, and the voice agents running in production at Zencia AI.

Languages

PythonJavaScriptC++SQL

LLM Engineering

Agentic AIPrompt EngineeringRAGLangChainLangGraphLlamaIndexMCPTool CallingStructured OutputsLangSmith

Models

GeminiClaudeGPT-4.1LlamaQwenMistralWhisperBGE Embeddings

LLM Infra

vLLMLiteLLMOllamaLoRA / PEFTFine-tuningHF Transformers

RAG & Retrieval

FAISSChromaDBHybrid SearchBM25Vector SearchEmbeddingsSemantic SearchReranking

Voice AI

WhisperDeepgramWebRTCTwilioSpeech-to-TextText-to-SpeechVAD

Backend

FastAPIAsync APIsWebSocketsREST APIsAuthBackground JobsLoggingDocker

Deployment

GCP Cloud RunVercelHugging Face Spaces

Deep Learning

PyTorchCNNsVision TransformersEfficientNetTransfer Learning
Field notes / Current work

What he's building at Zencia AI.

Since August 2025, Saksham has been building production voice AI and agentic systems at Zencia AI — here's the actual stack behind that work.

01

Real-Time Voice Pipelines

Speech-to-Text (Whisper, Deepgram), LLM reasoning, Text-to-Speech, streamed over WebRTC and Twilio — with barge-in support and Voice Activity Detection for natural, low-latency conversation.

02

Neural TTS Adaptation

Adapting transformer architectures for speech generation by replacing text token embeddings with audio token representations.

03

Agentic AI

LangGraph-based systems with tool calling and structured outputs, integrated into Google Calendar, CRM systems, and databases.

04

Production Backends

Async FastAPI services, WebSockets, background jobs, and logging — deployed via Docker to GCP Cloud Run and Vercel.

38Public GitHub repos
69GitHub followers
11Hugging Face Spaces
1Package published on PyPI
Field evidence / Shipped work

From idea to shipped code.

Public, checkable projects under the Parthmax name — not just claimed on a page, but live on GitHub, Hugging Face, and PyPI.

01 Active Development

CodePilot

An autonomous AI engineering system that turns natural-language specs into full applications: repository-level understanding, architectural planning, RAG-based context retrieval, and self-correcting test/lint loops.

View on GitHub →
02 Live Demo

FALCON

A multi-stage fact-checking pipeline: extract the claim, cross-reference it against Wikipedia and live search, verify with an LLM, and generate an explainable confidence score.

Try on Hugging Face →
03 RAG System

DocuMind AI

Document intelligence over large PDF collections — BGE embeddings, FAISS and ChromaDB retrieval, BM25 and semantic reranking, orchestrated across Gemini, Claude, GPT-4.1, and Llama-family models.

View on GitHub →
04 Published on PyPI

parth-dl

A lightweight, zero-dependency Instagram media downloader CLI — no login, no API key. Built and maintained in pure Python.

View on PyPI →
05 CLI Tool

movie-box

A Python CLI to search, download, and stream movies and TV series — cyberpunk interactive shell, async API calls, subtitle support, and MPV/VLC streaming.

View on GitHub →
06 Curated List

100 Best GenAI Projects 2025

A curated collection of 100+ high-impact open-source LLM, RAG, and AI agent projects — actively maintained as the ecosystem moves.

View on GitHub →
Credentials / Track record

Proof, not just a portfolio.

Formal credentials, algorithmic rigor, and hands-on model-evaluation experience behind the applied AI engineering work.

Algorithms

LeetCode

350+ Data Structures & Algorithms problems solved, strengthening the algorithmic problem-solving behind production system design.

View Profile →
Certifications

AI Agents & MCP

  • AI Agents — Hugging Face
  • Advanced Model Context Protocol — Anthropic
  • Data Analytics — Deloitte Australia
Education

IIIT Lucknow

M.Sc. Artificial Intelligence & Machine Learning, Aug 2024 – Jun 2026 (completed). B.Sc. Mathematics, Dr. Ram Manohar Lohia Avadh University, 2021 – 2024.

The story continues

Follow the journey.