AS

Forward deployed engineer at Future Path

Alok
Sharma

I take AI from a customer's messy workflow to something that runs in production, and I stay until it holds up.

  1. Map the workSit with the team and trace how a task really moves.
  2. Find the constraintLocate the step that costs the most time or accuracy.
  3. Decide if an LLM belongsSometimes the answer is a query, not a model.
  4. Ship itA working prototype in days, then hardened with guardrails.
  5. MeasureEvals and observability so it keeps working after I leave.

Work

Future Path

June 2026 to now

Forward Deployed Engineer

  • Embedded with customer teams to scope where AI fits in their operations, then build and deploy it inside their systems.
  • Own delivery end to end: discovery, prototype, production rollout and the evaluation loop that keeps it accurate.
  • Turn what I learn in the field into reusable patterns and product feedback for the core platform.

Trademo

Jan 2026 to May 2026

Backend / AI Engineer

  • Owned internal AI initiatives across analytics and operations, validating whether an LLM was the right intervention before building.
  • Shipped LLM and retrieval workflows that removed repetitive manual data processing, using Python services with n8n orchestration.
  • Designed guardrails for LLM features: input validation, structured output checks, retries with fallback models and access-scoped data.
  • Built Django/DRF services processing 100k+ records a day on Kafka and Celery + Redis. Cut API latency by 35% and added New Relic and ELK observability.

ANCKR

Oct 2025 to Dec 2025, remote

Full Stack Engineer

  • Built AI-driven CRM workflows for sales teams, joining Salesforce APIs and objects with LLM features, semantic search and automation.
  • Developed REST APIs, auth and access controls, React dashboards and embedding-based search across CRM records with a US team.

Deloitte USI

Feb 2024 to Oct 2025

Software Developer / AI Engineer

  • Built an evaluation loop for an enterprise RAG platform that raised answer accuracy from 20% to 95%.
  • Replaced manual document search across large internal knowledge bases with semantic retrieval on GPT, LangChain and Pinecone.
  • Scoped problems directly with business stakeholders, defined MVPs and delivered iteratively.
  • Shipped Django + React apps with AI built into existing workflows, deployed on Docker, Kubernetes and CI/CD.

Projects

Forge-AI

An autonomous coding platform. Give it a task and a GitHub repo, and six agents (product manager, architect, database, backend, frontend, devops) produce a plan, the code and optionally a pull request, streamed live to a pipeline dashboard. It's where I test agent reliability, retries, memory and human-in-the-loop control.

LangGraph-style orchestration, tool use, state management

Code Reviewer

A local-first CLI that reviews your git changes before you commit or open a PR.

Python Tutor Web

A step-by-step visualizer that shows Python code executing line by line.

LeetCode Sync

A Chrome extension that pushes each accepted LeetCode solution into a GitHub repo, one folder per problem.

WGANs for brain-computer interfaces

A Wasserstein GAN that augments EEG data, with CNN models for motor-imagery classification.

Skills

AI and LLMs
Prompt engineering, RAG pipelines, agents and tool use, multi-agent orchestration, evaluation frameworks, LangChain, LangGraph, Claude, GPT, Gemini, open-source models
Guardrails
Input validation, output filtering, prompt injection defence, access controls, data leakage prevention
Integration
Salesforce, n8n, REST APIs, webhooks, enterprise data models, Kafka, Celery, Redis
Engineering
Python, TypeScript, SQL, Django/DRF, FastAPI, Node.js, React, Next.js, PostgreSQL, MongoDB, Elasticsearch, Pinecone
Ops
AWS, Docker, Kubernetes, CI/CD, New Relic, ELK

Education

B.Tech in Computer Science, Indian Institute of Information Technology, Sonepat. 2019 to 2023, CGPA 9.33.

Coursera: Machine Learning (Stanford) and Neural Networks and Deep Learning (DeepLearning.AI), both 97.25%.

Have a workflow that should be automated?

Tell me how the work moves today. I'll tell you honestly whether AI belongs in it.