Engineering @ SigIQ · 3x Founding Engineer · AI agents, backend systems, and developer tooling.

agentpack · Portfolio · LinkedIn · Email · X

I work at the intersection of AI agents, backend/platform engineering, and practical product delivery. Across founding-team environments and 0-to-10 product work, my focus has been turning messy real-world workflows into reliable software: context engines for coding agents, scalable APIs, deployment automation, data-backed product surfaces, and infrastructure that engineers can operate with confidence.

name: Vishal Sharma
location: Gurgaon, India
role: Engineering @ SigIQ / 3x Founding Engineer
focus:
  - AI agents and developer tooling
  - Backend and platform systems
  - Distributed workflows and real-time infrastructure
  - RAG, MCP, context engineering, and evaluation
current_signal:
  - Building agentpack, a local context engine for AI coding agents
  - Shipping backend, AI infrastructure, deployment automation, and production workflows
  - Built scalable systems from scratch across startup environments
  - 2x hackathon winner
engineering_style:
  - correctness first
  - simple ownership boundaries
  - observable systems
  - practical delivery under ambiguity

• Engineering @ SigIQ: backend, AI infrastructure, payments, CI/CD, observability, and deployment automation for PadhAI product surfaces.
• 3x founding-engineer profile: repeatedly worked in ambiguous early-stage environments where engineering needs to connect product judgment, delivery, and systems ownership.
• Startup and enterprise product engineering: built client-facing platforms across learning, skill intelligence, commerce, sustainability/logistics, analytics, and goal-tracking systems.
• Open-source and community background: GSoC @ XFCE, Microsoft Student Partner, founder of FreshlyBuilt, and public GitHub work across AI tooling, backend APIs, analytics, and developer workflows.
• Builder history: 2x hackathon winner with older work across web platforms, machine learning experiments, BI dashboards, REST APIs, and startup/community projects.

• AI-agent infrastructure: local-first context routing, task-aware repo analysis, compact context packs, MCP integrations, and evaluation loops for AI coding agents.
• Scalable backend services: production APIs, async workflows, deployment automation, payment integrations, and service boundaries that remain debuggable as systems grow.
• Real-time and event-driven systems: WebSocket/SSE services, background job orchestration, retry-aware workflows, and operationally visible data flows.
• RAG and semantic systems: retrieval workflows, semantic search, context selection, prompt/runtime grounding, and evaluation strategies for LLM applications.
• Data and analytics products: dashboards, product intelligence surfaces, metrics pipelines, and systems that make decisions easier for teams.
• Reliability and cost-aware platforms: systems designed around latency, correctness, observability, rollback safety, and practical cloud cost control.

Problem: AI coding agents waste time and tokens rediscovering large repositories before making useful changes.
Architecture / stack: Python CLI, npm wrapper, MCP integrations, static repo analysis, task-aware file ranking, context packing, benchmark tooling, CI, docs, and local-first security/privacy design.
Why it is interesting: AgentPack treats context selection as an engineering system: token-budget aware, local-first, benchmarked, reproducible, and designed for real coding workflows across Claude Code, Codex, Cursor, Windsurf, MCP tools, and CI jobs.
Proof points: published PyPI/npm packages, quickstart/install docs, CI workflow, test suite, benchmark methodology, security policy, privacy/data-flow docs, and reproducible public-suite benchmark targets.
Links: Repo · PyPI · npm · Docs · Benchmarks

Problem: Production failures need more than alerting; teams need repeatable remediation workflows that connect runtime context, code context, fixes, review, and post-deploy learning.
Architecture / stack: AI-agent workflow orchestration, GitHub issue/PR loops, runtime context collection, CI/CD verification, deployment feedback, observability signals, and model-agnostic remediation flows.
Why it is interesting: Model-agnostic remediation pipeline that gathers evidence, opens issue-backed PRs, reviews fixes, tracks post-deploy behavior, and improves from merged outcomes.
Links: Private/internal project.

Problem: Scaled learning products need reliable backend systems, AI workflows, payments, deployment automation, and observability behind user-facing education experiences.
Architecture / stack: Python/Django, Node.js, PostgreSQL, Redis, AWS, CI/CD, payment integrations, deployment automation, observability, and AI-infrastructure workflows.
Why it is interesting: Combines product delivery with backend reliability: deployment automation, AI infrastructure, payment flows, operational visibility, and fast iteration across a live product surface.
Links: Private/company work.

Problem: Enterprise teams need structured learning, skill visibility, goal tracking, and operational workflows across large client-facing environments.
Architecture / stack: Backend services, LMS aggregation, skill-intelligence workflows, Jira-style goal tracking, APIs, data models, and client-facing product delivery.
Why it is interesting: Strong enterprise-systems signal: requirements ambiguity, multi-stakeholder delivery, product workflows, platform thinking, and backend design for operational users.
Links: Private/company work.

Problem: Discovery-led commerce products need fast product surfaces, reliable catalogue workflows, user-facing shopping experiences, and data-backed iteration across brands, products, and customer journeys.
Architecture / stack: Product engineering across commerce workflows, catalogue/product data, APIs, frontend/backend surfaces, operational tooling, and startup delivery loops.
Why it is interesting: Shows applied product engineering in a consumer-commerce environment where speed, reliability, experimentation, and clean product flows directly affect conversion and user experience.
Links: Private/company work. Public company context: furrl.ai

Problem: Sustainability-focused D2C logistics products need software foundations for packaging workflows, brand operations, tracking, and repeatable execution in an early-stage environment.
Architecture / stack: Founding-engineer product development across backend workflows, operational systems, dashboards/internal tools, and early product infrastructure.
Why it is interesting: Strong 0-to-1 signal: building software where product assumptions, operations, logistics, and engineering systems have to evolve together.
Links: Private/company work.

Problem: LLM applications need more than chat logs; engineering teams need traceable inference events, redaction, latency/error metrics, context checkpoints, and evaluation feedback loops.
Architecture / stack: FastAPI, React, Python SDK, SSE streaming, Redis Streams, Postgres, Docker Compose, Kubernetes manifests, provider adapters, eval fixtures.
Why it is interesting: Shows full-stack AI infrastructure depth: streaming chat, token-budgeted context recall, sensitive-data redaction, ingestion workers, DLQ replay, observability dashboards, and deterministic agent-harness telemetry.
Links: Repo

Problem: Early-stage products and communities need fast execution across frontend, backend, content workflows, analytics, and growth loops.
Architecture / stack: Web platforms, Django/API development, BI dashboards, Power BI/Grafana-style analytics, SEO operations, ML/NLP experiments, and community tooling.
Why it is interesting: Shows long-running builder range: founder work at FreshlyBuilt, analytics/API work for product teams, hackathon-winning projects, technical-content/community work, and earlier ML/chatbot experiments.
Links: Portfolio · GitHub Pages repo

• Design for correctness first, then scale the parts that evidence says are hot.
• Prefer simple systems with explicit ownership boundaries and boring failure modes.
• Optimize for debuggability: logs, traces, metrics, replayable cases, and clear operational runbooks.
• Treat latency, reliability, and cloud cost as product features, not cleanup tasks.
• Build systems that other engineers can safely extend without needing hidden context.
• Make AI workflows measurable: retrieval quality, token cost, tool-call waste, regression suites, and production feedback.

• AI coding agents and local context engines
• MCP-based developer workflows
• RAG evaluation and context-quality benchmarks
• Distributed backend systems and event-driven architecture
• Real-time infrastructure with WebSocket/SSE patterns
• Observability, incident remediation, and production feedback loops

• Senior backend / platform engineering roles
• AI infrastructure and developer tooling work
• Founding engineer opportunities
• Open-source collaboration around AI agents, context engineering, and backend reliability

• LinkedIn: linkedin.com/in/vishal-sharma-gbpecdelhi
• Email: vishalsharma.gbpecdelhi@gmail.com
• Portfolio: vishal2612200.github.io
• X: @humanlifevishal