In talks with Abhishek Sachar
In a tech landscape overflowing with AI buzzwords and automation promises, there’s a growing distinction between tools that simply “do” and those that truly “understand.” Enter Abhishek Sachar, an engineer who’s quietly shaping the next evolution of AI tooling—not by replacing human intent, but by aligning with it.
“I never wanted an AI that writes code for me. I wanted one that collaborates with me.”
That philosophy has defined Abhishek’s journey into AI system design. With over a decade of engineering leadership behind him, Abhishek has transitioned from building scalable backend systems and real-time infrastructure to developing intelligent agents powered by cutting-edge AI protocols. His latest work involves the implementation of AI Assistants connected through the Model Context Protocol (MCP)—a framework that allows agents to reason over real-time data across development environments, content repositories, and team tools.
Engineering with a Second Brain
The AI agent Abhishek built internally is not your usual chatbot or copilot. It’s more like a context-aware collaborator, deeply integrated into his engineering workflow.
It helps him:
● Review code across multiple repositories
● Auto-manage JIRA boards and sprint tickets
● Analyze team performance via Git activity, velocity, and PR quality
● Summarize engineering health metrics every Friday
● Assist in planning architecture refactors by querying past documentation
“It’s my engineering mirror. I trust it to bring up anomalies in logic, to summarize pull request quality trends, or even nudge me about developer morale. It saves hours—so I can spend more time thinking about product, not process.”
At the core of this evolution lies MCP (Model Context Protocol)—an open standard recently proposed to bridge the AI reasoning gap. Modern AI models, even the most advanced LLMs, are often stuck in isolated environments. They can write fluent English or code, but lack meaningful access to context like company docs, code history, or task boards. MCP addresses that fragmentation.
By creating secure, modular, and standardized pipelines to tools like GitHub, Slack, Google Drive, or even internal databases, MCP turns AI agents from passive responders into informed collaborators. For example, instead of asking your assistant “what’s the latest on X bug,” your AI can already know the JIRA status, related PRs, and who last worked on it—because it’s subscribed to that context through MCP.
Bridging Builders with AI, Creatively
While engineering productivity was his first playground, Abhishek soon expanded the agent’s scope. Using his internal MCP-powered agent framework, he recently built a set of automated content agents for his startup’s digital operations. These agents:
● Auto-generate weekly product updates for email newsletters
● Turn Notion-based marketing briefs into polished blog drafts
● Handle basic customer queries based on internal documentation and onboarding guides
● Track the performance of published content and suggest repurposing ideas
“I wanted to build something that empowers a five-person team to function like a fifty-person one—without losing voice or culture. That’s what agentic design lets you do.”
His vision isn’t about human replacement, but human acceleration. In fact, Abhishek credits his early success with agents to deep respect for human workflows.
“Every agent I built had a human-in-the-loop moment—either at review, approval, or creative generation. Agents aren’t here to erase roles. They’re here to eliminate friction.”
Innovation Born Out of Constraint
Like most meaningful breakthroughs, Abhishek’s entry into agentic AI wasn’t fueled by luxury—but by necessity.
“When you’re leading tech in a resource-constrained team, and you’re wearing six hats, you have to find leverage. I didn’t start building agents for fun—I built them because I was exhausted.”
Instead of hiring more people or outsourcing core ops, he went all-in on AI infrastructure. The result: a modular agent design that plugs into any environment using open-source standards and retrieves just the right amount of context to assist, not overwhelm.
The model connects to active sources (think: Slack, GitHub, JIRA), processes context through a minimal prompt layer, and delivers outputs that are actionable, auditable, and explainable.
“Explainability was key. If my agent says a PR is bad, I need to know why. If it flags burnout signals in commit patterns, I want traceable rationale. Not just AI guesses.”
Looking Ahead: AI That Feels Less Robotic
For the future, Abhishek sees AI agents evolving from reactive bots to predictive teammates. With protocols like MCP maturing, the dream is that anyone—designer, PM, marketer, or founder—can spin up a custom, context-rich assistant without engineering complexity.
He’s also interested in exploring agent swarms—multiple specialized agents talking to each other, managing workflows in parallel, and reducing human cognitive load to just what matters.
“We don’t need one superintelligence. We need 100 micro-intelligences, each focused, each collaborative, and each respectful of human agency.”
Advice for Builders
For engineers and founders stepping into this world, Abhishek offers one crisp takeaway:
“Start small. Don’t build for demos. Build to solve your real pain. An agent that saves you five hours a week is more powerful than one that dazzles but dies in production.”
His second piece of advice?
“Understand your data topology. AI agents are only as good as the context they see. If your systems are siloed, your agents will be blind.”
To keep up with Abhishek’s explorations and upcoming agentic frameworks, you can follow his LinkedIn where he shares early prototypes and insights. He’s also open to collaborating with early-stage startups experimenting with AI integrations.
In an age of AI abundance, Abhishek Sachar is quietly reminding the industry that the best agents aren’t built to impress—they’re built to empower.
