Abhishek Mukherjee

Software Developer, Vedant Asset Limited

2023 – Present, Ranchi, Jharkhand

• Building and maintaining mutual fund and insurance platforms using Next.js, Node.js, PostgreSQL, and AWS.
• Developed investment workflows covering onboarding, KYC, fund discovery, SIP creation, and portfolio tracking for 12,000+ users.
• Improved Lighthouse performance score from 48 to 91 through image optimization, code splitting, and caching strategies.
• Built backend APIs handling 40,000+ daily requests with p95 latency under 180 ms.
• Owned SEO implementation across investment and insurance products, increasing organic traffic from 6,000 to 28,000 monthly visits.
• Automated deployment workflows using CI/CD pipelines, reducing deployment time from 40 minutes to 6 minutes.

AskZen — askzen.in

• Fine-tuned Vicuna 7B on NCERT Class 6–12 textbooks using LoRA techniques.
• Built a teacher portal that generates lesson plans, quizzes, and curriculum material automatically.
• Serves 2,400+ students and 180+ teachers.
• Stack: Vicuna 7B, LoRA, Next.js, Python.

AI Tutor with RAG and Memory — Jharkhand Government Project

• Built a retrieval system combining textbook knowledge and conversation memory on every query.
• Stored historical conversations as vector embeddings in ChromaDB for personalized responses.
• Developed Flask APIs and retrieval pipelines supporting web and CLI interfaces.
• Achieved p95 retrieval latency under 120 ms across 18,000+ queries.
• Stack: Flask, ChromaDB, SentenceTransformers. GitHub

Basukinath Multi-Camera Surveillance System — Jharkhand Government Project, 2025

• Built a GPU-accelerated face recognition pipeline processing 1,553 comparisons per second.
• Improved recognition accuracy from 95% to 99.2% by replacing DeepFace with InsightFace.
• Added ROI-based detection and batch inference, doubling throughput across 16 camera feeds.
• Reduced manual verification workload by 50%.
• Stack: Python, InsightFace, GPU. GitHub

AEO Writer — Open Source — aeo-writer.vercel.app / GitHub

• Built an AI-powered SEO research and content generation platform.
• Supports competitor URL analysis, keyword discovery, article generation, and background processing jobs.
• Generated 3,200+ articles using OpenAI, Gemini, OpenRouter, Tavily, Redis, and Inngest.
• Stack: Next.js 16, TypeScript, Supabase, Redis.

Crackbase — crackbase.in

• AI learning platform focused on students from Tier 2 and Tier 3 cities.
• Built roadmaps, job discovery tools, interview preparation modules, and AI-powered learning workflows.
• Supports 4,500+ users.
• Stack: Next.js, Express.js, Supabase, PostgreSQL.

ATS Resume Builder — resume.crackbase.in / GitHub

• Open-source ATS resume builder that generates recruiter-friendly PDF resumes.
• Integrated AI-assisted content generation using Gemini API.
• Generated 6,800+ resumes.
• Stack: Next.js 16, TypeScript, Supabase, jsPDF.

PranSaathi

Next.js, TypeScript, Python, AI model fine-tuning, PLpgSQL, CSS

Saw government hospitals dealing with serious queue problems and built an entire app to manage it intelligently. Implemented the ABDM API, wrote my own algorithm to handle complex patient queues gracefully, and built an AI model that predicts virus and flu outbreaks 4 weeks in advance at a district level. Went out on my own, met the ABDM Joint Director of Jharkhand, got rejected due to unclear compliance but a great learning. Access restricted: pransaathi.vercel.app. Need access, mail mukherjeeabhishek207@gmail.com

SEO Freelancing

Generated 42K backlinks for Loopworm, an insect protein startup, boosting its Google ranking to top 5 in its niche. Developed a WordPress site for The Thoughts Company, increasing traffic by 60%. Created an expansion strategy for 3Bigs, driving 50% traffic growth in Middle Eastern markets. Built and optimized websites using JavaScript, WordPress, and Next.js, implemented SEO best practices, managed backlink profiles, and analyzed metrics using Google Analytics, delivering 30-50% performance improvements.

Research

LLM-based clinical decision support system for oral cancer treatment pathways in Indian healthcare settings.

Writing

Engineering articles covering RAG systems, async AI workloads, retrieval pipelines, and computer vision.

Teaching

YouTube channel focused on practical AI engineering and software development.

Building a Voice AI for Mutual Funds, and the Browser Speech API Wall I Hit

July 8, 2026

Implementing Leader Election in kvraft (M1)

July 2, 2026

Tracking Down a "Workspace Not Found" Bug in Dify

June 28, 2026

Building a Distributed Key-Value Store on a From-Scratch Raft

June 27, 2026

Building a Zero-Cost Voice Assistant on Free-Tier Infrastructure

June 27, 2026

Fixing 403 Errors While Keeping AI Costs Low

January 20, 2026

What Breaks First in Async AI Systems

January 8, 2026

A System Design Change I Didn't Plan For

January 7, 2026

From DeepFace to InsightFace: How We Took Face Recognition Accuracy from 95% to 99%

August 10, 2025

Building a Real-Time Face Recognition Engine with GPU Acceleration

July 20, 2025